• Skip to primary navigation
  • Skip to main content
GrowthPath Partners LLC

GrowthPath Partners LLC

Empowering Purpose-Driven Growth

  • Engagements
  • Speaking
  • Newsletters
  • About
  • Contact
  • Show Search
Hide Search

case study

How Box Turned 67 AI Opportunities Into a Clearer Plan for What to Build Next

Liza Adams · October 1, 2026 ·

How Box turned 67 AI opportunities into a clearer plan for what to build next, case study graphic

Sixty-seven AI opportunities is a good problem to have. But once the list gets that long, you need a way to decide what deserves attention first.

Box is the secure content layer supplying context to any AI agent, serving more than 120,000 paying customers. Its marketing organization already had real AI momentum. Teams were experimenting, and marketing had a growing set of agents.

Our discovery work with the marketing organization identified 67 possible AI use cases. Before we could prioritize them, we first had to understand how the work actually happened across people, processes, and systems.

From there, we needed to make three decisions for each one: What should happen to the work? Is it worth pursuing now? And how much should AI own?

Those questions became the heart of the blueprint we built with Box.

Quick Take

The blueprint turned 67 possibilities into a sequenced plan the team could act on:

  • More than 30 conversations across Marketing, SDR, Operations, and IT were translated into documented workflows, 8 core insights, 15 jobs to be done, and 67 possible AI use cases.
  • The Sorting Machine process helped us decide what the work needed before choosing the technology.
  • Impact, effort, and risk helped us decide what to do now and what dependencies to address first.
  • The Trust Ladder helped us decide how much of each workflow AI should handle and where human judgment should stay involved.
  • The resulting blueprint sequenced opportunities by maturity and dependencies. Box is now turning it into internal timelines, business cases, and project plans.
  • Reimagine the work first, then the roles, then the org. Let the work tell you what needs to change before you redraw the org chart.
From discovery to decisions: 30+ conversations, mapped workflows, 8 insights, 15 jobs to be done, 67 use cases, 3 lenses

Watch or Listen Instead

To support different learning styles and time constraints, this newsletter will also be available as a 4-min cinematic AI video explainer and a 20-min AI podcast with two AI hosts. Both were made with Google’s Gemini Notebook and personally reviewed for accuracy and responsible AI use.

Map the Work as a System

My business partner, Megan Cabrera, led the day-to-day discovery and blueprint work with Box. She ran stakeholder interviews, mapped workflows, developed the AI blueprint, and drafted the recommendations. I provided strategic direction, developed the frameworks, guided prioritization, and reviewed the work as it evolved.

Megan spoke with more than 30 people across Marketing, SDR, Operations, and IT to understand how the work supporting marketing actually happened. How did a campaign move from idea to execution? Where were the handoffs? Which systems held the information? Who owned each step? Where did manual coordination slow things down?

The interviews were only the first step. Megan translated what we heard into documented workflows so we could see the work as a system. Marketing is creative and cross-functional, and much of that work isn’t always documented end to end. Mapping the workflows made the handoffs, dependencies, ownership, and opportunities for change visible.

From there, Megan identified 8 core insights and 15 jobs to be done, each tied to an outcome the team needed to achieve, then broke them into 67 specific AI use cases. For example, planning and launching a campaign might include AI use cases for drafting the brief, identifying audiences, or analyzing past performance.

That order was deliberate. The workflow gives you the context to decide whether the work should move faster, change, stop, or become something new before you decide where AI belongs.

For Tricia Gellman, CMO of Box, that systems view was a big part of the value:

Headshot of Tricia Gellman, CMO of Box
Tricia Gellman, CMO of Box

“One of the most important parts of this work was translating what we heard in the interviews into actual workflows so we could assess the system. Once you can see how the work really happens, you can decide what to accelerate, what to change, and where AI can add value.”

That systems view gave us something concrete to take into the next step: evaluating the 67 opportunities.

Three Questions Helped Us Sort 67 Opportunities

Once the workflows were mapped and the 67 opportunities identified, we used three lenses, each answering a different question.

1. What should happen to the work?

The first was the Sorting Machine, which asks what a workflow needs before deciding how AI should be used. Work can take one of five paths:

  • Speed it up when the work is useful but takes too long or too much manual effort.
  • Sharpen it when the outcome is right but the process can work better.
  • Rethink it when the outcome is still useful but the process is not effective or not working.
  • Build something new when AI makes work practical that was previously too hard, slow, or expensive with an all-human team.
  • Stop doing it when the work no longer creates enough value to justify the time.

The point is to decide what the work needs before deciding what to build. I introduced the Sorting Machine and Trust Ladder in Think Before You Build: A Practical System for Human+AI Teams. That newsletter goes deeper on both frameworks. The Box engagement gave us a chance to apply them across a much broader set of work.

The Sorting Machine: speed it up, sharpen it, rethink it, build something new, or stop doing it

I’m especially interested in the concept of “Build something new” because AI removes constraints that influenced how we designed work in the first place. That can mean reading every sales call, continually refreshing competitive intelligence, tailoring insights to each account, or running scenarios before a decision. Work like this was often not done because it was too slow, difficult, or expensive with people alone.

I went deeper on those opportunities in 5 Reimagined Workflows: Work We Skipped That AI Now Does.

2. Is It Worth Doing Now?

The Sorting Machine tells you what kind of change the work needs, but not whether to pursue it now. So we evaluated each opportunity across impact, effort, and risk:

  • Impact – potential business benefit, including breadth, efficiency, revenue, and quality.
  • Effort – readiness of the tools, data, and process; the systems that need to connect; and the functions that need to be involved.
  • Risk – sensitivity of the information and output, exposure if something goes wrong, and how much autonomy AI would have.

Megan used AI to work through the volume while making the final call:

Headshot of Megan Cabrera, GTM Strategy and Operations Advisor
Megan Cabrera, GTM Strategy and Operations Advisor, Revenue Engine Labs

“With 67 use cases, AI helped us get through the volume, but I never let the scoring stand on its own. I asked it to justify every rating, checked those ratings against what we heard in discovery, and changed them when they didn’t hold up. AI made the analysis faster. The context still came from the people doing the work.”

A strong AI idea can still move later in the sequence. The data may need work, ownership may need to become clearer, or the process may need to change before AI is added.

The visual below shows a representative example of an impact, effort, and risk matrix. It is not Box’s actual matrix, use cases, or scores.

Sample impact vs. effort and risk matrix for AI opportunities

That helped us sequence 67 possibilities into a roadmap instead of treating them as 67 separate projects.

3. How Much of the Work Should AI Own?

The third lens was the Trust Ladder, which helps determine how much autonomy AI should have in a workflow. That depends on the work, the risk of a bad output, the quality of the inputs and instructions, and how easily a person can catch an error.

There are four levels:

  1. A person leads and AI assists. The person makes the decisions while AI helps with suggestions, drafts, or analysis.
  2. AI creates the first output and a person reviews it. Nothing gets used until a person has checked it.
  3. AI runs the workflow and a person spot-checks it. The work is proven enough that every output no longer needs review, but a person still owns quality.
  4. AI operates autonomously. This is reserved for well-proven, low-risk work where routine human review is no longer needed.

The Box blueprint combined the Sorting Machine, impact, effort, and risk, and the Trust Ladder into one way to evaluate each opportunity.

The same AI platform can operate at different levels depending on the job. It might help a marketer think through positioning at Level 1, create drafts for review at Level 2, and handle a proven, low-risk workflow with much more independence elsewhere.

The Trust Ladder treats autonomy as a decision about the work rather than a feature of the tool.

Three questions: what should happen to the work, is it worth doing now, and how much should AI own

The Roadmap Came After Those Decisions

Once we worked through the opportunities, the harder dependencies became clearer:

  • Data needed to be better connected.
  • Ownership needed to be clearer.
  • Some workflows needed to change.
  • Teams depended on shared systems and knowledge.

Those dependencies drove the roadmap, from foundational projects to data and workflow improvements to individual agents, orchestration across agents, and new ways of working.

An agent inherits the environment you give it. Inconsistent information produces inconsistent output. Unclear ownership stays unclear. And if the workflow itself needs to change, automating it can make the wrong process faster. Some of the most useful AI work happens before anyone builds the AI.

The Human Work Changes Too

Changing a workflow can also change who does the work, who owns decisions, where expertise is needed, and which handoffs disappear. People may need new skills. Leaders may need to redraw ownership. Teams may need shared definitions and outcomes. And someone still has to decide where human judgment adds the most value and where AI can take on more responsibility.

That’s why I believe we should reimagine the work first, then the roles, then the org. I wouldn’t start by redrawing the org chart. The work tells you which roles need to change and what the organization needs to support.

Box’s roadmap included enablement, training, governance, and evolving roles alongside the technology and workflow work.

For me, this is People First. AI Forward. in practice: every AI decision gets a people lens while we keep pushing toward work that wasn’t practical before.

I wrote more about that in AI Is Making the Human Side of Work Harder to Avoid. As AI removes some of the technical constraints around work, people still have to work through trust, ownership, incentives, roles, and decision rights.

Now Box Is Putting the Blueprint Into Action

The final blueprint gave Box a sequenced set of use cases and the dependencies behind them. Box is now building internal timelines, business cases, and project plans around those choices.

The team has a clearer way to decide what to pursue, what needs to happen first, and how much AI should own.

If Your AI Opportunity List Is Getting Long

A growing list of AI ideas can be a good sign. The harder part is deciding what deserves attention now.

Before you turn that list into a roadmap, ask:

  • What should happen to the work?
  • Is it worth doing now?
  • How much should AI own?

Those three questions helped us turn 67 possibilities into a plan Box could act on.


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.

80% Meeting Rates, 3x Reply Rates: A Real-Life AI Growth Engine Built Around the Customer

Liza Adams · February 5, 2026 ·

AI-native growth engine infographic: the five-stage AI growth flywheel and revenue results
Inspired by Sean McCaffrey, Sr. Dir of Strategy & Operations at Cin7

Around 98% of your website visitors leave without taking action. If that many people walked into a physical store, looked at products, checked out price tags, and walked out, you’d fix it tomorrow. Sean McCaffrey at Cin7 decided to fix it online. He built a five-stage AI growth engine designed around how customers actually buy, not around internal workflows. Today, Sean and his team are seeing 80% meeting booking rates, 3x higher email replies, 10% lower acquisition costs, and more human conversations.

Quick Take

This newsletter features how one strategic marketing operations leader built a connected AI system that captures demand in the moment, recovers interest that would otherwise disappear, learns from what closes deals, and feeds those insights back into every customer touchpoint.

Key takeaways:

  • AI creates growth when it’s wired into your revenue system around customer needs, not bolted on as isolated tools to do old work faster.
  • The right tech stack depends on your situation. This story shows what’s possible, not what to copy.
  • If we simply train AI to do old work faster, we unintentionally create a case for automating humans out. But when we use AI to reimagine work that wasn’t possible before, work that drives growth, humans become more essential. AI handles the repetitive work so humans can have better conversations.

AI Video Explainer and AI Podcast Versions of This Newsletter

To support different learning styles, this newsletter is available as an 8-min AI video explainer and a 15-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.

The Shift

Your buyers are researching in communities, asking peers, and using ChatGPT, Gemini, Perplexity, and Google AI Mode to evaluate vendors before they ever reach your site. They avoid forms. They expect relevant responses the moment they do engage, and if you don’t deliver, they leave without telling you why.

Sean and his team decided to build around the customer instead.

One Team’s Response

Cin7 is an inventory management company serving product sellers. Sean McCaffrey, Senior Director of Marketing Strategy and Operations, joined three and a half years ago. I worked with the team early in their AI journey in 2024 and have watched them build this over time.

His starting point wasn’t “what AI tools should we buy?” It was “how do we actually meet customers where they are?”

The answer became what Sean calls an AI-native growth engine. It captures demand when buyers are paying attention, recovers the ones who leave before converting, learns from what actually closes deals, and feeds those insights back into every customer touchpoint.

The flywheel has five stages. Each one exists because of a specific customer behavior. Sean and his team didn’t build all five at once. They started with one, proved value, then expanded. You can do the same.

AI-Native Growth Engine video explainer thumbnail

Stage 1: Engage in the Moment

Buyers don’t fill out forms like they used to, but they do respond when something feels relevant.

Sean’s team uses Qualified, an AI SDR that talks to website visitors in real-time. More on the specific tools later. This story is about the approach, not a vendor recommendation. It pulls firmographic data through a Clearbit integration, knows what pages they’ve seen, and adapts the conversation based on industry.

A visitor from a food and beverage company doesn’t get a generic “How can I help you?” They get asked about expiration date tracking and inventory spoilage. The conversation starts with their likely problem, not a blank slate.

Of visitors who engage and meet their criteria, 80% book meetings compared to 50% through traditional forms.

The AI isn’t replacing human conversations. It’s making sure qualified buyers actually get to a human instead of bouncing.

Headshot of Sean McCaffrey, Senior Director of Marketing Strategy & Operations at Cin7
Sean McCaffrey, Senior Director of Marketing Strategy & Operations

“Before, reps spent 30 minutes asking ‘what do you guys do?’ Now they walk in with full context.”

Stage 2: Recover What Would Otherwise Disappear

Most of those visitors aren’t gone forever. Sean’s team built a way to find them. Vector identifies who visited from the US. Clay adds context and confirms they’re a good fit. Together, they flag the people who looked seriously but didn’t reach out.

These aren’t cold prospects. They’re people who showed interest and match the profile. They just weren’t ready yet.

Stage 3: Re-engage with Context

The follow-up isn’t generic. Those contacts flow back into Qualified, which sends emails on behalf of the human SDRs. The emails are informed by what the system knows: which pages they visited, what their company does, what problems they’re likely facing.

Sean’s team draws a clear line here: lead with helpfulness, not stalking. The email doesn’t say “I saw you on our pricing page.” It offers a point of view on a problem worth solving. That’s the line between creepy and valuable.

The AI handles the repetitive work. Humans step in for the conversations worth having.

Reply rates are 3x higher than before.

Stage 4: Learn from What Closes

Closed-won calls recorded in Gong feed into AirOps, where the team extracts what actually worked: objections that came up, language buyers used, triggers that created urgency, competitive positioning that won.

The team uses those insights to update ad copy, landing pages, email sequences, and the AI SDR’s conversation flows. When “fear of stockouts during peak season” keeps showing up in winning deals, the team puts that same language into ads and outreach. The system learns from every win.

Stage 5: Multiply Without Burnout

Producing one piece of content and moving on leaves value on the table.

The team built a hub-and-spoke workflow in AirOps. One webinar becomes blog posts, LinkedIn content, email sequences, ad options, and sales talk tracks. A human reviews and approves the outline before full generation. Brand guidelines keep everything on voice.

The content is grounded in themes and pain points the team knows resonate because they’ve heard them in customer conversations.

Hub-and-spoke workflow turning one webinar into blog posts, LinkedIn content, emails, ads and sales talk tracks

What It Actually Took

This didn’t happen overnight.

  • Foundation first – The first year was making sure HubSpot and Salesforce were properly integrated and data was clean. You can’t build AI systems on messy data. Nobody wants to hear that, but skipping it breaks everything that comes after.
  • Content Cleanup – The AI SDR gave wrong answers in the first month. Sean traced it back to outdated help center content. The AI was doing its job. The content was the problem. They spent 30-45 days reading chat transcripts and cleaning up the knowledge base before expanding.
  • Controlled testing – Every tool ran in a sandbox before going live. Sean and his team earned leadership trust by bringing results, not just ideas. He added his CRO to the Slack channel showing real-time visitor alerts. Within a week, the CRO was on board.

Enterprise teams will face longer timelines with additional compliance, security, and procurement reviews. The principles still apply. Start with foundation and controlled testing, then expand.

The Results

Year over year, the team saw 10% lower customer acquisition cost, 20% lower cost per lead, 80% meeting rates through the AI SDR versus 50% through forms, and 3x higher reply rates.

The numbers tell part of the story. Here’s what changed day to day in the reps’ work according to Sean:

“They know the prospect’s business, their pain points, what content they engaged with. Discovery calls actually discover something. The AI handles repetitive admin work and analysis so humans can have better conversations.”

Before You Build Your Version

Sean’s specific tools solved his specific problems. Yours will differ. The underlying principles hold across industries, even when the tools don’t.

What I’ve seen across the teams I advise:

  1. Max out what you’re already paying for. Sean’s team uses HubSpot, Salesforce, Clearbit, and Gong as core infrastructure before reaching for anything specialized. Most CRM, marketing automation, and analytics platforms already have AI built in. Your existing platforms are already connected to your data and known by your team. Getting full value from that investment is faster to adopt and often takes teams further than expected.
  2. For the gaps, decide: build, buy, or keep it human-led. Sean’s system uses Qualified, Vector, Clay, and AirOps. These solved specific problems his existing stack and AI platforms alone couldn’t.

If you’re evaluating a build: start with a platform like ChatGPT, Claude, Gemini, or Copilot. Cin7 uses ChatGPT for everything from research and analytics to strategy. Build only for what’s truly unique to how you work. The demo is quick. Shipping and maintaining it is the real cost. Know what it costs to run at scale.

If you’re evaluating a buy: vet the vendor carefully. Many of these tools won’t be around in 18 months. Look at data trust, workflow fit, and whether they can keep up. Do trials. Stay flexible.

And some decisions are best kept human-led for now.

Your stack will look different. It depends on your goals, team, skills, and budget. Know which decision you’re making before you make it. Sometimes the answer is what you already have. Sometimes it’s a specialized platform. Sometimes it’s custom. Sometimes it’s a person.

Note: This isn’t a vendor recommendation. It’s a look at what becomes possible when you design AI around customer needs instead of internal workflows.

Framework: start with what you have, then decide whether to build, buy or keep it human-led for the gaps

The Bigger Point

Using AI well means more than using ChatGPT, Copilot, Gemini, or Claude. AI can be wired into your entire revenue system, but only if you design it that way.

Sean and his team aren’t doing old work faster. They’re doing work that wasn’t possible before: engaging every visitor in real-time, recovering demand that would have disappeared, learning from closed deals at scale, creating content informed by what actually resonates.

If we simply train AI to do old work faster, we unintentionally create a case for automating humans out. But when we use AI to reimagine work that wasn’t possible before, work that drives growth, humans become more essential.

Customers expect you to know who they are, meet them where they are, and pick up the conversation where it left off. In the AI era, company size matters less. The advantage goes to teams that learn how to listen, respond, and improve continuously.

The money follows when you earn trust. We now have the tools to do that better than ever.

Your Next Steps

Sean’s flywheel is one version of what’s possible. You don’t have to build all five stages to start seeing value.

  1. Where is demand leaking? Where do high-intent visitors go when they don’t convert? If you don’t know, that’s the first problem to solve.
  2. Are you building for growth or productivity? Are you using AI to reimagine work around customers, or only to do old work faster? The answer determines whether humans become more essential or more optional.

I’d love to hear what you’re building. Reply to this email or share on LinkedIn. The trailblazers in this community are figuring this out together, and what you learn helps others move forward.


Want practical AI insights like this delivered every two weeks? Subscribe to the Practical AI in Go-to-Market newsletter.

For applied learning: Our applied AI workshops offer both strategic sessions (use cases and roadmaps) and hands-on building (create AI teammates during the workshop). You’ll leave with either a clear plan or working solutions.

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’d love to connect.

Copyright © 2026 · GrowthPath Partners LLC · Log in

  • LinkedIn