AI workflows

Your Data Problem Is Smaller Than You Think. Your AI Payoff Is Bigger.

Liza Adams · August 20, 2026 ·

A data readiness project comes with a cost and a timeline. What it unblocks is often unclear, so it gets funded on faith or it doesn’t get funded at all.

That’s where I started with a marketing team. I could have spent the session showing them how to fill out their channel performance decks faster, and automating that would have saved real hours. We went a different direction.

I asked what it would take for the same workflow to tell them which blog posts compete for the same readers, which persona they’re under-serving, and what to test next. Then a second AI agent would read across all three channels (blog, gated assets, and webinar) and challenge the plan before anyone spent budget on it.

Everyone wanted that version. So I asked one question: how do you tag your content today, by topic, by persona, or by funnel stage?

There was very little tagging at all. The closest thing was a product code sitting inside the blog page address, like the “exp” in /blog/exp/cut-expense-report- cycle-time/. AI had nothing it could filter on, group by, or read as a category.

That team isn’t unusual. In Dun & Bradstreet’s latest AI Momentum Survey of 10,000 businesses across 32 countries, only 6 percent said their enterprise data is fully ready to support AI at scale. In the same survey, more than three quarters reported some measurable return from AI anyway.

That question uncovered our first data gaps. Working out what the better version needed uncovered the rest. We built it so we could test it and see the outputs. Based on the outputs, we refined, and that’s how we confirmed data was the issue.

Quick Take

The usual advice is to get your data in order first, then use AI. But there’s another way.

Ask AI for the more valuable version of the job, and it will tell you what your data is missing and how each piece would be used.

Here’s what happened when we tried it with one marketing team.

Key takeaways:

Build first with dummy data. Set a higher goal, reimagine the workflow, and find out what AI can and can’t do before you invest in getting the real data.
Filling out a report faster works fine on messy data. The problem is when you ask for analysis by persona, product, or funnel stage.
You can find some gaps before you build anything. Ask your team how the data behind that workflow gets categorized today. If you get different answers, you’ve found one already.
Sometimes the answer is that the data was never collected and never will be. Better to learn that in one working session than after months of data work.
Be clear about which demos are running on dummy data. Otherwise the disappointment gets pinned on AI.
Fix one gap first, the one that blocked what you wanted most. A small win gives you something working to expand from.

There’s a prompt at the end you can run on your own workflow.

The four steps we took, in order.

Four steps: decide the output, ask what it takes, build on dummy data, close one gap

Decide the more valuable output you want, ask AI what that output requires, build it on dummy data where the real data isn’t there, then start by closing the one gap that blocked what you wanted most.

Watch Or Listen Instead

To cater to different learning styles and time constraints, this newsletter is also available as a 9-min AI video explainer and an 11-min AI podcast with two AI hosts. Both were made with Gemini Notebook and personally reviewed for accuracy and responsible AI use.

Speed Hides the Problem

A workflow that pulls three channel performance exports, formats them, and populates a template saves hours whether or not anything is tagged well.

The data problem shows up when you want analysis instead of a finished report: results by persona, recommendations by funnel stage, or a coach that challenges your plan. All of that asks AI to sort and compare by category. If nothing in the file shows those categories, you get observations back. This blog did well, that one didn’t. It’s accurate but not as helpful.

That’s the difference between faster, better, and different, and each level asks more of your data. (That framing is reframe four in The 8 Reframes That Change How GTM Teams See AI.)

What each level of AI work (faster, better, different) asks of your data — What each level asks of your data.

Here is the same channel report with and without those data gaps.

Sample blog channel report with no tags, goals or definitions

With no tags, no goals, and no agreed definitions, you get a list of your most- read posts and no judgment on whether any number is good, because there’s nothing to judge it against.

The same channel report after closing the data gaps

Once we closed the data gaps, the same workflow could tell you the expense- management cluster aimed at finance buyers is your fastest-growing topic, and the one page in it that matches reader intent keeps people reading for 38 seconds. Your highest-traffic page loses them in one second and is tagged to the wrong product. Rewrite that page and publish more in that cluster.

The first version stops at one bad page. The second tells you what to write next, for which buyer, and where the gaps are.

Both run on synthetic data, including the tags that made the second one possible.

Build It With Dummy Data

We didn’t stop to fix anything first, because we didn’t yet know what needed fixing. The team generated dummy files that kept the same structure as their real exports, with fake names and numbers, so we weren’t processing sensitive data in the experiment.

Check your company’s responsible AI policies before you start. If the data you’d need is customer-identifying or regulated, find out what your policy allows you to collect and where you’re allowed to process it.

If your policy allows it, you’ll eventually build with real data. If it doesn’t, dummy files let you keep going, and you’ll know from the start that closing the gap may mean redacting fields, replacing identifiers with codes, or asking for aggregated numbers only.

Then we built on the dummy files.

A channel analyst agent reads a single export and produces a completed KPI report with insights and recommendations.
A coach agent reads across three channels and pushes back on the plan.

Channel analyst and team coach agent workflows with human review gates

The dummy fields standing in for data we didn’t have became the requirements list.

The usual path vs. building first with dummy data

Eight Things We Were Missing

We decided the output we wanted, then asked AI what it would take to produce it. Recommendations by category meant tags had to exist for topic, persona, and funnel stage. A trend meant a date field, which two of the three exports didn’t have. Judging whether a number was good meant a goal for the period, a benchmark to compare against, a baseline from your own history, and a target. Knowing what a page actually accomplished meant conversion tracking. And the funnel stage definitions had to be written down where the AI could use them.

Each answer we wanted turned into a requirement, eight in all, and the dummy files filled those in.

They picked the tags first, on one channel, because that was the gap blocking the recommendations they wanted.

Eight things we were missing and what we couldn't do without them

Tell Them What They’re Seeing

There’s a version of this that goes poorly. You show a team an impressive output built on dummy data, everyone gets excited to try it on their real data, it doesn’t come close, and they decide AI is overhyped.

So I pre-empted this at the start of the session. The AI is doing what it can with what it has.

It stops people from blaming the tool for a weak answer, and it turns the excitement into a list of data requirements for consideration.

Fix the Gap that Blocked What You Wanted

Close one gap, not the full list. Pick whatever blocked the output you most wanted. If it’s recommendations by persona, tag one channel. If leadership asks for trends every quarter, add a date field to that one export.

Getting all of it in place can run six months, stall the rest of the process, and be out of date by the time it’s done. Closing one gap gets you started and leaves you with something working you can expand from.

Dun & Bradstreet’s chief strategy officer made the same point. A pilot can run on the data you have. Scaling across the systems people depend on is where AI- ready data becomes the constraint. We can run the pilot with one workflow on one channel, then address the data gaps needed to scale.

Ask your analytics team what data you have and you get a list of everything. Ask whether persona can become a field on the blog export and you get a yes or a no. Go to whoever owns the data with the second question.

Julie Zadow, Growth Marketing Advisor at The Riverside Company, sees this from the investor side, across a portfolio of marketing leaders.

Julie Zadow, Growth Marketing Advisor at The Riverside Company

“I work across a portfolio of CMOs, and I’ve seen the same data challenge play out repeatedly: someone walks into budget planning and says the data needs work, but nobody knows exactly what they’re approving. If a CMO walks in with four specific fields and the workflow each one unblocks, that’s a completely different conversation.”

When you ask that question, answers come back in one of four ways.

The field already exists and nobody was using it.
Adding it is a small piece of work rather than a project.
It isn’t there yet, so you start collecting it and the workflow includes it later.
It will never exist, so you change what the workflow produces and build around what you have.

Any of those is better than funding a data readiness project and finding out at the end. Until people see the output they want, they keep using the same data they always have.

Justin Parnell, my business partner and a forward deployed engineer who specializes in AI automation and agents, built a territory planner for a client whose data warehouse wasn’t ready. The agent had to rank accounts and map each rep’s stops for the day.

Justin Parnell, Forward Deployed Engineer and Founder of Justin GPT

“Instead of waiting, we built on synthetic data matching the structure the warehouse would eventually have. The first maps the agent drew put pins in the middle of lakes, and chasing that told us precisely what the real data had to look like, down to storing zip codes as text so a zip like 02134 doesn’t lose its leading zero and land you in the wrong state.

The data team never had to tell us they hadn’t collected it. We found it first, and came to them with a tested checklist instead of a vague ask.”

Some answers depend on someone else’s timeline. If the field lives in a system another team or a vendor owns, getting it means asking them and waiting. Find that out before you commit to a date. (On sorting which workflows deserve the effort, see Think Before You Build.)

Think big, start small, move fast.

Run This on Your Own Workflow

Open your AI tool of choice and give it this prompt:

I want to build an AI workflow that does this job: [describe it in plain language, including the analysis and recommendations you want, not just the report].

Here are the inputs I can give it today: [list your exports and documents, with column headers if you have them].

Tell me what fields have to exist in those inputs and what has to change about how they’re structured. Then tell me what other sources I’d need and what definitions we’d have to agree on first. After that, what you could produce today versus what I’d have to add, and which one to fix first, and why.

Then flag anything on that list we can’t get retroactively, so I know what to stop chasing.

Pick one workflow you wish AI could run for you. Write down every field, definition, and target it would need, then pull the export you’d feed it and compare. That list is your starting point, and it’s shorter than fixing your whole data foundation. Close a few of those gaps and the same workflow starts telling you what to do rather than what happened.

The Practical AI in Go-to-Market newsletter shares learnings and insights in using AI responsibly. Subscribe today and let’s learn together on this AI journey.

For applied learning: Explore our applied AI workshops, offering both strategic sessions to shift mindsets through practical AI uses and hands-on sessions to build human+AI workflows. You’ll leave with either a clear plan or working solutions. Office hours provide follow-on support to keep momentum going.

For ongoing advisory: I work with leaders as a strategic advisor through their AI transformation, providing guidance, frameworks, and a sounding board as they build. If that’s what you need, let’s talk.

For team transformation: I partner with GTM leaders to move from scattered AI adoption to an operating model where human and AI teammates work together across workflows.

For speaking: Here are virtual and in-person events where I’ve covered a variety of AI topics. I’ve also keynoted at many conferences and corporate events.

If any of this resonated, I’d love to hear from you. Reply to this email, find me on LinkedIn, or come say hi at an event.

5 Reimagined Workflows: Work We Skipped That AI Agents Now Do

Liza Adams · July 9, 2026 ·

5 Reimagined Workflows: campaign synthesis, competitive intel, sales enablement, AI search, customer signals

There’s a whole category of work your team’s not doing. Reading every sales call. Refreshing battlecards every week. Tailoring the message to each deal. You skipped it because it was slow, manual, or expensive to do by hand.

Some of the most important work also needed you to look outward and keep looking, at the whole market, at each account’s world, at what buyers say across hundreds of calls. Your team looks at what it can see, which is its own data. Not from a lack of effort, but because one team can only do so much. AI can look outward, and do it continuously.

That made sense before AI. For a lot of this work, the reason you skipped it is gone. The question worth asking about any task you skipped: is it still hard, or just historically hard?

Quick Take

The GTM workflows I see being reimagined aren’t revolutionary. They’re practical, teams are running them today, and they’re changing how the work gets done. They’re the work we always wanted to do, but it was more than an all-human team could handle.

What I share here comes from working with teams just a few months into agentic workflows becoming more real and accessible for GTM teams. These teams lean AI-forward, so take it as early signal, not a perspective on the whole market. And the point is never AI. It’s the work AI makes possible.

Key takeaways:

AI now makes more work doable. But we have to fix the habit of not doing this work.
The best inputs are already in your systems: your calls, usage data, support tickets, win/loss notes.
Human teams look at what they can see, which is their own data. But AI can look outward too, at the market, at each customer’s world, at industries adjacent to yours, and bring back ideas with the reasoning.
Five agentic workflow examples: build one plan from insights across all signals, watch the market and inform strategy, tailor the proof and practice to each deal, see how you show up when buyers ask AI, and understand how your customers are changing inside and out.
AI handles what it does well: collection, synthesis, analysis, and execution at scale. Humans keep judgment and the final call.

Watch Or Listen Instead

To cater to different learning styles and time constraints, this newsletter is also available as a 6-min cinematic AI video explainer and a 15-min AI podcast with two AI hosts. Both were made with Google’s NotebookLM and personally reviewed for accuracy and responsible AI use.

5 Most Reimagined GTM Workflows

Each one runs as an agentic workflow. It moves through the steps on its own, pauses where your judgment is needed, uses other AI and tools to handle parts of the work, and delivers finished files you can use. You guide it, make the key decisions, and review the output.

1. Build one plan from every signal (campaign synthesis)

Today, with an all-human team:

Teams coordinate in meetings and alignment calls, but something still gets missed.
When the plan changes, it’s hard to reflect those changes wherever needed.
The plan leans on one or two sources and misses what cuts across all of them, including what’s shifting in the market and in your buyers’ world.

Campaign synthesis workflow: what you feed AI, the human+AI workflow, and what you get

With human+AI:

The plan is built from every source at once, internal signals and the market, and stays current when inputs change.
Positioning applies across the buying committee because it reflects all of it
The buyer gets one consistent story across every touchpoint, which builds trust earlier in the cycle.

The value is the synthesis, not just the drafts. A change made in the plan stage is one fix, not multiple ones, that gets applied throughout quickly.

Christine Matu, VP of Product Marketing at Blue J, shared her thoughts after a team hands-on AI workshop.

Christine Matu, VP of Product Marketing at Blue J

“We used to pull personas, messaging, customer call transcripts and win/loss interviews from various sources and stitch them together every campaign. We want to ensure we’re always build from what customers actually say instead of the loudest opinion in the room, and that was challenging to do manually. Now it’s within reach. I’m looking forward to campaigns that resonate and reflect what our customers care about most, in real time.”

Blue J hands-on AI workshop: messaging assessment output from Claude Cowork — Blue J Hands-On AI Workshop – Messaging Assessment Output from Claude Cowork

2. Watch the market and turn it into strategy (competitive intel)

Today, with an all-human team:

Someone owns competitive intel, but watching every competitor’s site, pricing, reviews, and your own win/loss data is more than one person can keep up with, so it happens in bursts.
Battlecards get built once and then go out of date. By the time a competitor’s new claim reaches your reps, a few deals are already lost to it.

Competitive intel workflow: what you feed AI, the human+AI workflow, and what you get

With human+AI:

AI watches continuously across the market, adjacent categories, and your own calls.
Reps get current battlecards with both sides, offense and defense, tied to the objections buyers are raising now.
Leaders see the change early. When a competitor repositions or a new objection spreads across calls, you can adjust messaging and packaging before it impacts pipeline.

The most insightful input is your own call recordings and customer reviews. What a buyer says about a competitor is more honest than what the competitor says about itself.

3. Tailor the proof and practice to each deal (sales enablement)

Today, with an all-human team:

Enablement builds one deck and one pitch for everyone, because tailoring the proof to each deal takes more time than any team has. So reps get generic material and adapt it on their own.
The buyer’s own costs and savings sit in a spreadsheet a rep may or may not build well under deal pressure.
Reps practice on live prospects. The hard objection gets worked out in a real deal and they learn too late, typically in the loss review.

Sales enablement workflow: what you feed AI, the human+AI workflow, and what you get

With human+AI:

Reps get proof built for the deal they’re working, an ROI case using the buyer’s own numbers, and messaging informed by what buyers in your category actually struggle with, not just your own talking points.
The tools are self-serve and always current, so every rep gets the same quality of prep, whether it’s the day before a big call or ten minutes before.
Reps practice the hard conversation before the call, against a simulated buyer who pushes back, so the first real objection isn’t the first time they’ve answered it.

The simulator isn’t scripted. Each run is a live conversation, so the rep faces a different buyer and/or situation every time.

4. See how you show up when buyers ask AI (AI search)

Today, with old habits:

We optimized for keywords and rankings. That still helps AI find us, but ranking for a term and being the one AI recommends are not the same thing.
We led with features and capabilities. AI can’t tell from a feature list who we serve best or which problems we’re the right choice for, so it fills the gap by guessing, and you may not like the guess.
We treated our own website as the source of truth. AI also reads review sites, communities, and third parties, and leans on what others say about us, not only what we say about ourselves.

AI search workflow: what you feed AI, the human+AI workflow, and what you get

With human+AI:

You see how AI describes and recommends you across the major engines, whether you show up, how you’re framed, whether you’re on the short list, and which sources were cited.
You find the gaps that lose deals like a wrong description, a missing proof point, a competitor recommended over you, before more buyers form the same impression.
The fix is clarity. When you say plainly who you serve best and where you’re not a fit, AI routes the right buyers to you and steers the wrong-fit ones elsewhere. You win more of the deals you should win, and spend less time on the ones you can’t serve well.

One run is a snapshot but it’s not the full picture. Reliable signal takes many runs over time, and for some teams an AEO/GEO tool built for scale might be the better solution. I go deeper in When AI Judges Your Brand Before Humans Do.

5. Understand how customers are changing, inside and out (customer signals)

Today, with an all-human team:

The signs that an account is growing or at risk are split across your own systems, usage, tickets, engagement, call notes, and the customer’s own world, funding, acquisitions, leadership changes. No one watches across all of it.
CS has too many accounts to watch closely, so the quiet ones get noticed only near renewal or after a problem shows up.
The signal that matters most is often external and easy to miss. A customer being acquired is a churn risk. A customer acquiring is potentially an expansion opportunity. You usually find out too late to act on it.

Customer signals workflow: what you feed AI, the human+AI workflow, and what you get

With human+AI:

The workflow reads your internal signals and watches each account’s public external world, then scores the account for expansion or risk with the reasons behind it.
Your team sees which accounts need attention and why, early enough to act. That includes changes in the customer’s business you’d never see in your own data.
The customer gets the right support at the right moment, a check-in when they’re stuck, a conversation when their business is changing, an offer that fits where they actually are.

The insight that matters most is often outside your data. A customer being acquired can tell you more about churn risk than any usage trend.

This is also where silos cost you: the signal one team sees that another needs. AI reading across all of it is how fewer things fall through the handoffs. More on that in No Daylight.

The Outside-In Advantage

All five workflows share the outside-in advantage. The campaign reads the market alongside your own numbers. Competitive intel watches adjacent categories, where substitutes and new entrants tend to appear first. AI search shows you how the market already sees you. Customer signals look at public events in your customer’s world that usage data alone would miss. Even enablement gets better when it pulls in what buyers across the market struggle with.

A human still judges whether an outside idea applies. Some are useful, some aren’t. What changes is you have options you didn’t have before.

What’s Common Across Workflows

One thing was true of every workflow: A human makes the final call and is accountable.

The best workflows use human gates. You approve the plan before it goes forward. You review the output before it goes out. Quality checks run at key points.

At each gate, we instruct AI to show its work, not just hand over an answer: options with tradeoffs, confidence levels, risks flagged, what’s likely to land and what isn’t. That’s critical thinking built into the workflow. More on that in AI Is Only as Sharp as Your Questions.

Some workflows earn more autonomy over time. Once one proves itself, you can check the output instead of every step. Keep the gates where a mistake is costly or hard to fix. Let it run where the work is low-stakes and easy to undo.

It’s Only as Good as What You Feed It

These workflows are only as good as what you feed them and how you guide them. They need clean data from a single source of truth, clear instructions and context, the right access and permissions, human oversight on the judgment calls, and governance for safety, compliance, and cost. Someone has to own each one. And the outside-in research uses public signals only, news, filings, and announcements, not private surveillance.

But those are not reasons to wait. You don’t need all of it perfect to start. Pick one workflow, get the data and guardrails right for that one, and learn. A simple way to start: upload company-approved documents and data rather than connecting systems directly to AI.

Think big, start small, move fast.

Faster is the Floor

If your AI strategy is focused on efficiency alone, you’re building a case for fewer people. If it’s focused on reimagining work to grow the business, you’re building a case for why humans are essential.

Growing companies keep adding humans. Stagnant ones start asking why they need so many.

The difference is whether you use AI only to do old work faster, or also to do new work that wasn’t possible before.

Laura Heisman, Chief Marketing Officer of Dynatrace, got hands on keyboard with her leadership team instead of just directing the work.

Laura Heisman, CMO of Dynatrace

“This isn’t about using AI to write a better email. It’s about rethinking how we actually do things. The leaders who win this chapter are the ones who translate AI into better ways of working for their customers and teams.”

Dynatrace marketing leadership team — Dynatrace Marketing Leadership Team

There’s a timing piece to this. Today, reimagined work is an advantage. Over time, it becomes what customers expect, and the advantage fades. So the question is when you do this work, and who decides. The teams that move now get ahead of it. Wait, and you’re matching a standard someone else set.

You can build these in whatever AI tools your team already uses. They all run agentic workflows now.

If you want the system for deciding what to hand to AI and how much, I offer some guidance in Think Before You Build.

Digital twins help here too, used as thinking partners more than task-doers. I’ve gone deep on those lately. A Cox Automotive team and a Morningstar team each built their own.

Where to Start

Pick one task your team stopped doing because it was too hard, too slow, or too expensive to do at scale. Ask whether it’s still hard, or just historically hard. Most of the time now, it’s the second one. That’s your first workflow.

Reimagining from a blank page is hard. It gets easier once you’ve seen a few examples, because you stop asking “what could AI do” and start asking “what did we stop doing that’s worth another look.”

What’s the one task your team stopped doing that might be worth another look?

The Practical AI in Go-to-Market newsletter shares learnings and insights in using AI responsibly. Subscribe today and let’s learn together on this AI journey.

For team transformation: I partner with GTM leaders to move from scattered AI adoption to an operating model where human and AI teammates work together across workflows. See real examples: a lean GTM team’s step-by-step playbook and a global cybersecurity leader scaling to 150+ marketers with 57 AI teammates integrated into daily workflows.

For speaking: Here are virtual and in-person events where I’ve covered a variety of AI topics. I’ve also keynoted at many conferences and corporate events.

If any of this resonated, I’d love to hear from you. Reply to this email, find me on LinkedIn, or come say hi at an event.

You’re Building AI Teammates and Workflows. Do You Know What They Cost?

Liza Adams · April 16, 2026 ·

What GTM teams need to know about AI pricing, prototypes and production before they scale

Most of us are focused on building AI teammates and workflows right now, and we should be. The early results are faster execution, better output, time back in the day, and even new ways to work that weren’t possible before. When something is working, you keep building.

What many haven’t bumped into yet is what these workflows cost when they move beyond individual use. When AI was just a chatbot, the cost was simple. You asked, it answered, your subscription covered it. Now AI is doing work on its own. It’s running steps, calling tools, and making choices while we’re not watching. That change is what makes the cost question different, and GTM teams don’t have to figure it out from scratch.

For this one, I asked Megan Cabrera to think it through with me. Megan is a GTM Operations Strategist and Advisor who works on the operational side of AI adoption, including the cost questions most GTM teams haven’t hit yet. Her perspective influences a lot of what’s in this piece.

I wanted to write this before most of us need it. This is what we think GTM teams should know to scale. We want to elevate the conversation so that we can all learn together and from each other.

Quick Take

We think of AI subscriptions as a fixed monthly cost. For individual use and even for connected workflows within your subscription, it is. When workflows need to run at org scale, production reliability, or in front of customers, the pricing shifts to variable, usage-based models that many GTM teams haven’t encountered yet.

Engineering and product teams have been managing this for a while. Megan and I are bringing what they’ve learned to GTM.

What we’re seeing so far:

AI has three pricing tiers based on how and where your workflow runs: fixed subscription, variable API (Application Programming Interface) costs, and unpredictable volume when workflows are public-facing. This is like three tables at a restaurant, each with a different menu and a different level of visibility into what you’re spending.
Per-token costs are dropping over 90% by 2030, but total spend is rising because complex workflows consume 5 to 30x more tokens per task (Gartner, March 2026). That’s cheaper ingredients but a bigger total bill.
Cost is one of several evaluation criteria. Architecture readiness, ongoing ownership, skills and team readiness, time to production, and business risk all factor into whether you should build a custom workflow, buy a specialized tool, or keep the work human-led.
The gap between a working prototype and a production-ready workflow is bigger than many teams expect. A polished demo is not a finished product.
The governance and enablement structure to manage this evaluation probably already exists on your team, often in your operations function. But it needs a new competency: understanding token economics well enough to simulate workflow costs before deployment.
The good news is your technical counterparts likely have experience with usage-based pricing, whether from cloud or AI. Whoever handles enablement can bridge what they’ve learned to what GTM needs to know. We don’t have to figure this out from scratch.

AI Video Explainer and AI Podcast Versions of This Newsletter

To support different learning styles, this newsletter is available as a 7-min video explainer and a 20-min AI podcast with two AI hosts. I used Google’s NotebookLM to create these and personally reviewed them for accuracy and responsible AI use.

Three Tables at the Same Restaurant

I think about AI pricing as three tables at a restaurant. Each one has a different cost model and a different level of visibility into what you’re spending.

How AI pricing changes as you build: buffet, à la carte and market price

Table 1 – The Buffet

Your subscription is $20 to $200 a month depending on the plan, fixed price, and more capable than many teams realize. You can build AI teammates, connect workflows to your email, Slack, and Google Drive through built-in integrations, and set up tasks that run on their own. You can even chain Custom GPTs (AI teammates you build in ChatGPT for specific tasks, trained on your data) or build workflows in Cowork (Claude’s desktop tool for knowledge work), all without leaving your subscription.

You’ll hit usage limits before you hit a surprise bill. For most individual and team- level work, you can stay here longer than you think, and you should. (Helpful tips: Charlie Hills shares 8 ways how you can get 3X more from your Claude subscription.)

Table 2 – The À La Carte Menu

When a workflow outgrows what your subscription can handle (volume exceeds usage caps, or it needs real-time responsiveness or production-level reliability), you move to the API, which is the direct connection used to build AI into custom applications and systems like your CRM, your project management platform, or workflows that run on their own.

Every time an agent (an AI teammate that takes actions on its own, not just a chatbot you prompt) runs, it’s consuming tokens, making API calls, and potentially invoking external tools or search. Tokens are the small chunks of text AI reads and generates, roughly one word each.

Megan pointed out that a single agentic workflow might look simple on the surface, say an automated competitive intel brief that pulls from web search, synthesizes findings, and shares via a Slack summary. Behind the scenes that one workflow is making a series of AI calls, often 10 to 20 or more depending on how it’s built, each one consuming tokens. Multiply that by frequency and team size, and you’re looking at real spend.

Gartner’s March 2026 research backs this up. Per-token costs for large AI models are projected to drop over 90% by 2030, but agentic workflows use five to 30 times more tokens per task than a simple chatbot conversation. The price per call is going down. The number of calls per workflow is going up much faster. Cheaper ingredients, bigger total bill.

The AI cost paradox: cost per token down 90%, tokens per workflow up 5–30x, total bill rising

A tech executive I know connected AI to his email, project tracking, and task management through the API and let it run on autopilot. He let it run without spending limits, and within weeks the charges were piling up. He described it as walking out of Costco with three carts full of tokens. The technology worked exactly as designed. The bill was a different story.

Once a workflow is running at scale, here’s one way to make the cost concrete: think of it as CPM by workflow type. Every marketer already knows how to manage cost per thousand impressions in a media buy. Cost per thousand runs of a specific workflow is the same shape, with one important difference. Each workflow has its own cost profile. A simple lookup and a multi-step agentic workflow on the same product can differ by 100x. Track and manage them separately, the way you’d track separate ad campaigns.

Megan flags that this is avoidable. The AI companies all support spending caps and usage alerts. Set them before you turn anything on, and the Costco moment doesn’t happen.

Table 3 – Market Price

AI pricing as a restaurant metaphor: buffet, à la carte menu and market price

The table with no prices listed. When your AI workflow is customer-facing, prospect-facing, or open to the public, you’re on the API and you’ve lost visibility into volume. Every person who interacts with it runs up a tab. If the experience works well and gets popular, your success is directly tied to your cost. Every time AI handles a request, that processing costs something.

Most GTM teams have operated in a world with seat-based SaaS pricing. You buy 10 licenses, and you know what you’re getting. With AI-powered experiences, every interaction has a compute cost. More usage means a bigger bill, even if the number of users stays the same.

Megan also points to model choice as a major lever. The difference between running a workflow on a larger, more capable model versus a smaller, faster one (for example, Claude Opus versus Claude Haiku) can be 10x to 30x on the same task. Think of it the way you’d think about delegation. You wouldn’t put your most expensive person on a task an intern could do. Use the bigger model for high-value, low-volume work where reasoning quality is important. Route the high-volume repetitive work to the smaller, faster, cheaper model. It’s one of the biggest levers teams can use to control cost at this table.

Caps are even more important here, and they work differently. You’re capping not just your own usage but also what happens when external demand exceeds your budget. Does the experience degrade gracefully, queue up requests, or pause until the next billing cycle? Worth thinking through before you go.

Workflows like personalized buyer experiences or automated onboarding can create real competitive advantage. Don’t avoid Table 3. Just understand the pricing model.

Who Should Own the Tab

One question GTM teams need to answer is who pays the tab. When marketing spins up an agentic workflow on the API, does that hit the marketing budget? IT budget? A shared AI center of excellence that may or may not exist yet? Most organizations haven’t answered that question, which means the bill is either going unmonitored or heading toward a finance clawback conversation.

When budget ownership is ambiguous, agentic initiatives stall at the worst possible moment: after the proof of concept works, right before scale. The conversation to have is “who will own the cost when it grows, and what does that approval chain look like.” That conversation turns AI adoption from an experiment into something procurement and finance can support.

Megan Cabrera, GTM Operations Strategist & Advisor

“In a lot of orgs, marketing ops or rev ops is already tracking vendor spend and tool budgets. Where that’s true, AI workflow costs are the next thing to add to the list, and most teams haven’t started. You don’t need a new function to do it. You need your existing ops lead talking to tech peers.”

The Beautiful House on Sticks and Twigs

There’s another gap between prototype and production that isn’t about cost. It’s about whether the thing you built can function in a business context.

Vibe coding has made this more visible. If you haven’t heard the term, it means using AI to build software by describing what you want in plain language rather than writing code line by line. You vibe and AI codes. You can go from idea to working prototype in an afternoon. The result looks polished, it functions when you click around, and you can deploy it for people to use.

“Looks finished” and “is production-ready” are two very different things.

A head of demand gen I work with saw this across several teams. Someone builds an interactive ROI calculator, a competitive comparison tool, or a custom landing page in a day. The demo is impressive and leadership gets excited. Then someone asks whether the components are built in the CMS, whether the lead forms flow into Salesforce, whether there’s analytics tracking, or whether the content team can update anything without going into the custom HTML. The answer is usually no across the board.

She described it as building a beautiful house on a foundation of sticks and twigs. It looks like a house. You can walk through the rooms. There’s no plumbing, no electrical, and no foundation that will hold up in bad weather.

A beautiful house on sticks and a web page with no connections: prototypes without infrastructure

The prototype took an afternoon. Connecting it to your business systems, getting it through security review, configuring it for your environment, and testing it against real data is where the real time and budget go. AI has made the integration work faster (writing connectors and configuration code goes much quicker now), but the approvals, testing, and rollout still take days to weeks. Those timelines are mostly organizational, not technical.

Some teams discover that the cost and effort of making a prototype production- ready ends up being more than a purpose-built SaaS tool would have cost for that same job. Worth knowing before you commit to either path.

Build It, Buy It, or Keep It Human-Led

Megan and I have been working through an evaluation rubric for this. It covers cost structure (including what happens at scale), architecture readiness, ongoing ownership and flexibility, skills and team readiness, time to production, human role, and outcome and risk. It applies whether you’re running on the subscription or on the API. The cost row changes, but the rest of the evaluation applies either way.

This builds on the three-layer tech stack approach I’ve written about before: start with what you already have, then go deep on AI models, then add specialized tools where they solve a clear problem. The rubric below is the cost and readiness lens you layer on top of that decision. (Also refer to “Your AI Teammates Aren’t a Team Yet” and the Cin7 AI growth engine case study for more insights.)

This assumes your team is already reviewing for security and compliance. These are the additional criteria to layer in.

From prototype to production: build, buy or keep it human-led

Megan and I agree that the case for Buy is often stronger than it gets credit for. Your CRM, MAP, ABM tools, and workflow platforms are all embedding AI features quickly. For standard GTM motions like lead scoring, email personalization, and call analysis, the embedded AI in your existing stack is usually good enough, and it comes with a big advantage. It’s already connected to your data, your team knows the interface, and the cost is often bundled into your existing contract. Watch for the exceptions, but there’s value in the default.

Teams that rush into building agentic workflows can end up duplicating what they already have, then carry the maintenance burden when those tools need debugging, updating, or scaling.

There’s another angle to the Buy case. The companies offering specialized SaaS tools have people whose entire job is thinking about one problem. They’re talking to dozens of customers solving the same thing, watching what’s working, and shipping improvements. Your in-house build is competing for attention with six other priorities, and the workflow you launched six months ago is probably not getting the iteration cycles it needs. Focused beats distracted, even when distracted is more customized.

The case for Build is real, and narrower than it looks at first. Custom AI workflows win in the gaps between your tools, in the connective tissue and processes that are unique to how your organization operates. The value is in the logic itself, which is unique to how your business operates.

A competitive intelligence pipeline that pulls from specific sources, synthesizes against your positioning framework, and distributes to sales in a format they’ll actually use is a good example. Build for the 15 to 20% (range may vary by company) of workflows that are genuinely differentiated, that connect systems in ways your vendors don’t, or that create customer-facing experiences you can’t buy off the shelf.

Sri Srinivasan, Chief AI Officer at Hellman & Friedman, shared the guidance he gives their portfolio companies.

Sri Srinivasan, Chief AI Officer at Hellman & Friedman

“Before building custom AI, ask what makes you special. If AI is your product’s advantage, build it. If it’s just how you run your business, use existing tools. Many teams waste time building things that will soon be standard features. The real cost isn’t building—it’s maintaining and improving it over time. Most value comes from configuring what already exists, not creating something new.”

The Structure Probably Already Exists

If you’ve been reading my newsletters, you’ve seen me write about the enablement function: the group (or person) that handles governance, IT and legal coordination, system integrations, and cost management on behalf of builders. (For how this works in practice, see “AI Moved Fast: What Changed and What to Do Next” and the cybersecurity company case study.)

The model is democratized building, centralized enablement. Everyone closest to the work keeps building. A smaller team makes it safe and scalable. In some organizations that’s a dedicated team. In others it’s one person in ops who manages governance alongside everything else. The roles still apply. How they map depends on your team’s size and structure.

Many organizations already have some version of this: marketing ops, rev ops, or an AI governance council. Builders submit use cases. Legal weighs in on risk. IT weighs in on security. Builders then build the workflows.

What I’m finding in my own client work is that the economic lens is the piece usually missing. Cost governance for AI workflows is new territory because the cost model itself is new. The review structures teams use for security and compliance can extend to cover it, but the economic lens has to be added on purpose. When teams see the move from buffet to à la carte, the gap is obvious, and the conversation turns to whether finance and P&L owners should be in the review.

Charlie Treadwell, CMO at Elisity, leads the strategy and rolls up his sleeves on the build. He’s been working through this in real time and shared what his team is doing.

Charlie Treadwell, Chief Marketing Officer at Elisity

“Our finance team is working hard to rebuild our budget models. We used to plan around program spend plus headcount, that was the fixed part of the equation. Now there’s a new variable.

The easy case would be: AI makes someone 2x more productive, you cut a role, you spend that money on tokens. That’s not what we’re seeing. We’re getting more productivity out of the same people, but it’s costing more money, and it’s hard to predict.”

Kelly Herrell, a CEO, posted a response on LinkedIn that takes this further, connecting AI workflow costs directly to the profitability question the CEO cares about.

Megan’s framing on this is that GTM teams should treat AI consumption like usage-based infrastructure. Engineering teams have been managing usage- based pricing for a while, whether it’s cloud bills or AI inference costs (what it costs to run a model in production), and they know what visibility and controls this kind of pricing needs. You need active visibility into what’s running, how often, and what it costs per execution, or you’ll find your team facing a monthly invoice you didn’t expect.

If you’re adding the cost lens to your existing process, here are some questions to start with.

What does this workflow cost to run at production volume, not just at prototype scale?
Is there an existing tool, vendor, or human-led process that gets the same outcome? How does it compare across the evaluation rubric?
Is this connected to our business systems, or is it a standalone prototype that still needs integration work?
If this is customer-facing or public-facing, have we modeled what happens to cost and performance when volume spikes?
Who on the team will maintain this after it launches?

If the builder can’t answer all of these, that’s expected. That’s what the enablement function is for. Your marketing or rev ops lead who already tracks vendor spend and tool budgets is a natural owner, and they’ll want to connect with their technical counterparts to learn how usage-based AI costs work. Many engineering and product teams have been tracking inference costs, setting token budgets, and evaluating build-vs-buy trade-offs. The ones that haven’t yet are working through the same questions we are. Either way, we can learn from them instead of rediscovering it on our own.

What We’re Still Figuring Out

The evaluation rubric is a starting point, and we expect it to evolve as we all learn more. A few open questions we’re sitting with:

How do you compare truly different cost structures? A SaaS subscription is predictable but rigid. A custom AI workflow is flexible but variable. A human-led process comes with institutional knowledge you can’t replicate but doesn’t scale cheaply. We’re still learning how to evaluate these side by side in a way that’s useful.
How should this change as AI matures? A workflow that’s too expensive or too complex to run on the API today might make sense in 18 months. Do you build now and wait for the rest to catch up, or pilot at low volume and revisit later? Teams are making different calls depending on how strategic the workflow is.

If you’ve started running workflows through a cost lens, what’s on your rubric that isn’t on ours? We’ll share what comes back.

Know What You’re Ordering

AI is worth the investment. The people on your team who are experimenting and building are doing exactly the right thing. They need better visibility into what happens when a workflow moves from prototype to production, and a place to ask questions before the bill shows up.

The structure to manage this probably already exists on your team, whether that’s a dedicated enablement function, your ops lead, or someone who wears that hat part-time. Add the economic and readiness lens. Bring finance into the conversation. Learn from your technical counterparts, whether they’ve figured it out or are figuring it out with you. And actively monitor any live workflows.

This week, whoever handles enablement on your team should connect with engineering or product about how they’re managing usage-based infrastructure costs. Then pick one workflow your team is building or recently launched and run it through the evaluation rubric.

The Practical AI in Go-to-Market newsletter shares learnings and insights in using AI responsibly. Subscribe today and let’s learn together on this AI journey.

For applied learning: Explore our applied AI workshops, offering both strategic sessions (use cases and roadmaps) or hands-on building (create AI teammates during the workshop). You’ll leave with either a clear plan or working solutions. Office hours provide follow-on support to keep momentum going.

For speaking: Here are virtual and in-person events where I’ve covered a variety of AI topics. I’ve also keynoted at many organization and corporate-wide events.

If any of this resonated, I’d love to hear from you. Reply to this email, find me on LinkedIn, or come say hi at an event.

AI Moved Fast: What Changed and What to Do Next

Liza Adams · March 5, 2026 ·

Scale AI: democratized building, centralized enablement

You’ve seen the LinkedIn posts. Someone built an app in 20 minutes. Someone else automated a full content workflow over the weekend. What those posts don’t show: the enterprise security policies, the legal reviews, the IT constraints, the fact that half your team doesn’t have access to the same tools. AI capabilities jumped in the last eight weeks. Your team’s security policies, legal reviews, and IT constraints stayed right where they were.

Quick Take

AI platforms shipped more capability in the last eight weeks than in the previous six months. Autonomous agents, desktop AI assistants, enterprise workflow studios, all went live in a short timeframe. If you feel like you missed a month and everyone else already caught up, you didn’t and they haven’t.

The progression I’ve been writing about still holds: Tool to Teammate to Workflow. But what’s possible inside each stage changed fast enough that your mental model from December needs an update.

Key takeaways:

AI teammates now have agency. They plan, execute, and check in at decision points. The shift from “I ask, it answers” to “I delegate, it delivers” is real, and most teams haven’t caught up to it yet.
Self-contained workflows stay inside your AI platform. Anyone can build them today. System-connected workflows touch your CRM, HRIS, or marketing automation. Those need IT and Legal. That determines whether you can start now or need to loop in others first.
The gap between “what’s possible” and “what’s possible for your team” depends on constraints like security policies, legal reviews, tool access, and organizational complexity.
Not everyone needs to be orchestrating human+AI workflows now. Knowing where you are is more important than racing to the next stage.
There’s an interactive AI Evolution Framework guide at the end that helps you find your starting point, the right platforms, and whether you need IT involved.

AI Video Explainer and AI Podcast Versions of This Newsletter

To support different learning styles, this newsletter is available as an 6-min AI video explainer (see below) and a 19-min AI podcast with two AI hosts. If you haven’t seen these AIs in action, they’re worth a view. The tech is advancing in amazing ways. I used Google’s NotebookLM to create these and personally reviewed them for accuracy and responsible AI use.

Here’s What Actually Happened

Eight weeks that changed AI: platform launches from Anthropic, OpenAI, Google and others

That’s eight weeks. For context, Custom GPTs launched two years ago and most teams are still figuring those out.

If the people you work with seem confused, stressed, or frozen, this is why. The capabilities moved but the constraints didn’t. IT policies, security reviews, legal approvals, tool access, organizational complexity, these are all right where you left them. And nobody is mapping the difference between “what’s technically possible” and “what’s actually possible for your team, in your org, with your constraints.”

The Stages Haven’t Changed. What’s Inside Them Has.

If you’ve been reading this newsletter, you know the progression: Tool to Teammate to Workflow.

Stage 1 got better. When AI gives you a ‘good enough’ answer, you stop pushing. You stay at Stage 1 longer than you need to.

Stage 2 is where the real change happened. AI teammates now have agency. They plan, break work into steps, execute, and check in with you at decision points. The shift from “I ask, it answers” to “I delegate, it delivers” is real. They work inside your computer, navigating files, apps, and browsers to get the job done while you focus on something else.

Anthropic’s head of product described Cowork as enabling AI to “deliver polished, near-final work.” OpenAI’s CEO said you can go “from a clean sheet of paper to sophisticated software in a few hours.”

But the biggest change is at Stage 3. Workflows split into two distinct modes, and the line between them changes who needs to be involved, what you can build on your own, and where governance actually matters.

The interactive AI Evolution Framework guide maps each stage in detail, including platforms and who needs to be involved.

The AI progression: from tools to teammates to workflows

Self-Contained vs System-Connected Workflows

The most important distinction in this whole framework is one you won’t find in any AI product announcement. It’s the line between self-contained and system- connected workflows.

1) Self-contained workflows keep everything inside your conversation or uploaded files within your AI platform. You chain multiple AI teammates together. You feed the output of one into the next. You produce a complete sales toolkit, a campaign brief with supporting assets, a market analysis with recommendations, all from a single starting prompt.

No external system connections. No sensitive data flowing through APIs. You can build these today, by yourself, without permission from anyone.

2) System-connected workflows reach into your CRM, your HRIS, your marketing automation platform. They read and/or write sensitive data. They need connectors, security reviews, and often legal sign-off.

You’ll hear the term MCP (Model Context Protocol) more and more. It’s how AI connects to business systems, and it already has over 100 million monthly downloads. Anthropic and Microsoft both launched enterprise connector ecosystems. This infrastructure is moving fast.

These workflows need an enablement team, working with IT and legal, to build the connections. You bring the knowledge and expertise of the jobs to be done. They provide the infrastructure.

This helps answer the common question: “Can I build this myself, or do I need to loop in others first?”

For self-contained work, start. For system-connected work, partner with your enablement team.

Who Makes This Work at Scale?

When I say “enablement team,” I don’t mean a new department. I mean a function. It might live in marketing ops, rev ops, IT, or a small cross-functional group. Someone who understands how work flows across teams, knows how to build automations and integrations, and can work with IT and Legal on security. People with this combination are hard to find right now.

They’re responsible for:

Training people to build well from the start
Supporting them when they get stuck
Prioritizing which system connections to build now and which to hold
Setting up governance so the organization isn’t flying blind
Watching the costs, because usage-based pricing on automation tools, API calls, and connectors adds up fast once workflows go live

That last one catches people off guard. Most AI subscriptions are flat-rate but most workflow infrastructure is not. Few teams are budgeting for this yet. The training, support, governance, and cost management are what keep it from breaking.

Here’s what it looks like when you find these people. Some may be in your org already.

A cybersecurity company's human+AI marketing org with AI teammates — Human+ AI Transformation Case Study with Builders and Enablers

At a global cybersecurity company I work with, Marketing Ops stepped into this role. Everyone across marketing built their own AI teammates freely. Over 200 were created during experimentation. 57 are now in daily use. But when those teammates needed to connect to CRM, marketing automation, or internal repositories, Marketing Ops handled the integration and governance. They worked with Legal and IT to make sure the access was responsible and secure. Full case study here.

The model is democratized building, centralized enablement. Everyone builds. A smaller team makes it safe and scalable. This works because the people closest to the work know what to build. A product marketer knows what a positioning teammate should do. A field marketer knows what an event planning teammate needs. You don’t want to bottleneck that. But when those teammates need to connect to business systems, you need someone who knows how to do that without creating risk.

Dominic Abou-Jaoude is the Director of Marketing Data & Analytics at Sophos, sits on his company’s AI Committee, and built an internal triage system that helps people figure out when to bring in his team.

Dominic Abou-Jaoude, Director of Marketing Data & Analytics at Sophos

“The way I know when to step in is pretty straightforward. If someone is building something with their own files and their own thinking, they should just go do it. I want them to. That’s how they figure out what works.

The second a workflow needs to push or pull from a core system, like Salesforce, then I step in to help. Because one bad connection to a live system could create a stability or a security issue. My job is to work with our engineering and security teams to make sure the plumbing is secure and reliable so people can keep building on top of it.”

Kimber Spradlin, Chief Marketing Officer at Graylog, is coming at this from the other direction. Her team already has custom GPTs. Now she’s moving them toward orchestrated workflows using Claude.

Kimber Spradlin, Chief Marketing Officer at Graylog

“I built a drag-and-drop workflow designer for our team offsite inspired by Liza’s dynamic ‘AI teammate and workflow Ideator’ app. Vibe-coded the whole thing.

The reason I could design the system is because I’d already built the individual piece using custom GPTs. I knew what each step needed because I’d spent time using them one at a time first. Now we’re moving it to a production server and building the actual agents behind each step. We learned a lot from building and using custom GPTs . It was the reason I could see how the pieces fit together.”

Here’s the dynamic AI workflow app that sparked Kimber’s approach. It offers sample AI workflows and AI teammates for each one.

Dominic and Kimber are at different starting points in the same progression. Both are building.

Find Your Starting Point

Your situation is different from the person sitting next to you. Different tools, different constraints, different starting point. A CMO at a 5,000-person company and a VP of Sales at a 50-person startup are both reading this newsletter right now, and their next move isn’t the same.

I built an interactive AI Evolution Framework guide that helps you figure out yours. It has three tabs.

Start with the decision guide on the first tab. Pick what you’re trying to do, and it maps you to the right stage, the platforms that fit, and whether you need to involve your enablement team. Takes about 30 seconds.

The other two tabs go deeper. One breaks down what’s changed inside each stage, including the platforms available now and how to know where you are. The other shows the operating model: what you build, what your enablement team builds, and how the roles work together. The short video below walks through both.

Interactive AI Evolution Framework Guide

See the 90-sec video preview of the interactive guide below.

If you want the quick-reference version, the one-pager maps every platform to every stage across all four ecosystems (Google, Microsoft, Claude, OpenAI) on a single page. Keep it open when you’re deciding what to build next. Here’s the one-pager.

Platform guide: which AI tool to open for each kind of work

What Comes Next

The framework gives you a starting point. These are the decisions that come after.

How fast should teams move from Stage 2 to Stage 3? Some teams I work with are still building their first AI teammates. Others are ready to orchestrate. The temptation is to push everyone toward workflows, but rushing past the teammate stage means you miss the part where people learn to evaluate AI critically. That skill is what makes Stage 3 work.
Where does the enablement team role land in smaller organizations? At a large company, you have an ops team, IT resources, and legal review capacity. At a 50-person company, the person building AI teammates might also be the person managing governance. The roles still apply. The question is how they map when one person is wearing three hats.
Are Custom GPTs a dead end? Ethan Mollick said they are. I get why. AI is moving toward systems that orchestrate themselves, picking tools and skills without you directing every step. But the thinking behind the GPTs is important. The knowledge you put in, the way you learned what AI is good at and where it falls short, the judgment about when to trust the output, all of that carries forward. You can’t set up AI-orchestrated workflows well if you’ve never orchestrated them yourself.

Go Deeper

Your Turn

This is my best thinking right now. It will keep evolving as the platforms advance and as I learn with teams putting this into practice. I want to hear from you.

Where is your team in this progression?
Where are you stuck?
What would be most helpful to go deeper on?

I read every response, and your input directly shapes what I write next. The conversations that feed this newsletter go both ways. Always have.

The Practical AI in Go-to-Market newsletter shares learnings and insights in using AI responsibly. Subscribe today and let’s learn together on this AI journey.

For team transformation: See real examples—a lean GTM team’s step-by-step playbook and a global cybersecurity leader scaling to 150+ marketers with 57 AI teammates integrated into daily workflows.

For speaking: Here are virtual and in-person events where I’ve covered a variety of AI topics. I’ve also keynoted at many organization and corporate-wide events.

Whether through the newsletter, multimedia content, or in-person events, I hope to connect with you soon.