Dispatches from All In: The Real Work of AI Transformation

Walking through the doors of All In, Canada’s largest AI event, you sense something different about this moment in time. This isn’t the era of “Will AI matter?” anymore. The question has shifted entirely: How do we actually build and deploy AI at scale?

I’ve spent 4 years leading marketing for AI product companies selling to enterprises in heavily regulated industries like construction and manufacturing. I’ve sat in whiteboarding sessions with founders who’ve lived the real-world problems they now solve: contractors who managed schedules for 30 years and manufacturing engineers who could spot a supply chain cascade three steps ahead. I’ve watched as some enterprises scaled AI adoption and others got stuck in pilot purgatory. The patterns resonate over time.

After connecting with founders and leaders at the ALL IN conference #TeamToronto reception hosted by the City of Toronto, and absorbing several keynote talks, what I heard echoed everything I learned from years in marketing with tech companies. The companies winning with AI aren’t the ones betting on breakthrough models. They’re the ones doing the unglamorous foundational work of fixing data, building governance, restructuring teams, and learning how to operate at the edge of what’s possible. They’re the ones who understand that marketing and selling AI to enterprise buyers requires patience, domain knowledge, and an obsession with solving real problems.

Here’s what ALL IN insights stood out for me, and why it validated everything I’ve learned about enterprise AI adoption in the real world.

Talk One: You Can’t Skip the Foundations

An ALL IN session entitled, “AI Readiness: Data Strategy, Talent, and the Rise of the CAIO,” pulled back the curtain on something most organizations aren’t ready to face: ambition alone doesn’t move the needle.

Panelists from Siemens, AWS, OpenText, and The Globe and Mail weren’t shy about naming the real barriers to AI transformation. They highlighted structural, rather than technical challenges:

• Fragmented data across legacy and modern systems
  
• Unclear ownership of AI strategy and data governance
  
• Weak governance frameworks that don’t scale beyond pilots
  
• Talent gaps at every level, especially strong AI architects who understand infrastructure

I’ve seen this exact bottleneck in enterprise AI adoption in construction and manufacturing. These industries run on systems built over decades—legacy ERP platforms, siloed project management tools, regulatory compliance databases. When an enterprise AI company tries to plug in without understanding why the data is fragmented, they fail. The fragmentation isn’t a bug; it’s a feature of how heavily regulated industries operate. Selling into these spaces means understanding the constraint before you can solve it.

What struck me most was the conversation about who owns AI strategy. There was agreement around the room that it can’t be owned by IT alone. It has to be a CEO-driven mandate, with the CAIO (Chief AI Officer) working as part of a trifecta with the CHRO and CFO. This isn’t bureaucracy—it’s alignment. When your CFO understands the ROI model and your HR leader gets the talent story, transformation actually happens.

As a marketing leader in this space, I learned this truth early: you’re not selling to IT. You’re selling to the business case. Enterprise adoption moves when the CFO sees time-to-value against a problem they already know costs them millions. When the CEO has mandate. When HR is prepared for the people implications.

The Data Cleansing Reality Check

One of the most practical takeaways: don’t wait for perfect data. The panelists emphasized that organizations need to put data cleansing practices in place at a consistent cadence, treating it as an ongoing operational function rather than a one-time fix. This reframes the problem. You’re not trying to achieve data perfection; you’re building muscle around continuous improvement.

This landed hard for me because it’s the message that finally got enterprises to move. In construction and manufacturing, I worked with companies sitting on years of messy data. They were waiting for “clean data” before they’d even pilot AI. The marketing shift that worked? Showing them that AI could work with messy data if they committed to the cleansing cadence. It wasn’t about achieving perfection first; it was about accepting reality and building the muscle. That reframing—from blocker to operational capability—is what moved deals from “maybe someday” to “let’s pilot this quarter.”

The Talent Trifecta Problem

Talent kept surfacing as the critical bottleneck. AWS noted that while it’s hard to attract top AI talent (they gravitate toward the “big ones”), the bigger challenge is empowering the entire organization through AI. This means building guardrails, providing support systems, and creating the infrastructure for people who aren’t AI experts to deploy systems effectively. It’s less about hiring AI PhDs and more about building an organization that can learn at scale.

This is where marketing messaging gets critical. In my work with AI companies in construction and manufacturing, I found that enterprises weren’t afraid of the technology—they were afraid of the people problem. Will this displace workers? Do we have the skills internally? Can we afford to retrain? The companies that scaled were transparent about this. They positioned AI as the tool that freed up your best people from rote work to do what they actually got hired for. A senior project manager shouldn’t be manually reconciling schedule conflicts across three systems. An engineer shouldn’t be typing data into spreadsheets. That’s the value story that resonated and drove adoption. It’s not “AI will replace your team.” It’s “AI lets your team do the work they should have been doing all along.”

Staying Ready in a Moving Target

One panelist posed a question that should keep CIOs and CEOs awake: Can a company ever truly be AI ready? Technology evolves weekly. The real question isn’t “Are we ready?” but rather, “How do we build an organization that can adapt as new capabilities emerge?”

AWS’s mantra captured this perfectly: “Live in 45.” The idea is to pick one fundamental way you want to change how your organization works—the transformation of the way you work—and execute on that in a 45-day cycle. It’s not about waiting for the perfect strategy; it’s about moving with intentionality.

Here’s where identifying the right customer matters tremendously. Not all enterprises are ready for this kind of agility. The ones that scale are the ones with leadership patience and appetite for transformation—but also with clear decision-making authority. In my experience marketing AI to construction and manufacturing companies, the ones who moved fastest weren’t necessarily the biggest. They were the ones with a champion executive—someone who’d fought battles to modernize the business before, who could rally their peers, and who’d stake their credibility on the change journey. Those customers don’t just adopt your product; they evangelize it. They become advocates internally and externally. Finding and prioritizing those ideal customers is worth exponentially more than chasing every prospect in the addressable market.

Talk Two: Agents Are Real, But They’re Not Magic

The second session, “Agents in Production: Turning Autonomy Into Results,” shifted focus from readiness to execution. And here’s what I learned: the hype about agentic AI is real, but the execution challenge is being vastly underestimated.

This talk grounded the conversation in concrete use cases. One panelist shared an example from supply chain management: customers were manually overseeing order changes that rippled through demand and supply plans. Now, agents access the right data, run scenario simulations autonomously, and present humans with decision points. The agent doesn’t make the final call; it does the heavy lifting and hands off to the human for judgment.

This pattern matters because it answers the most important question: Where does autonomy actually create value?

And here’s where something clicked for me that goes to the heart of marketing AI to enterprise buyers: this is exactly how you sell into regulated industries. In construction, you don’t automate the decision to change a schedule by three weeks—that decision affects labor, logistics, client commitments. What you automate is the analysis. The agent pulls scenario data, identifies cascading impacts, and presents options. The human—the project manager with 20 years of experience—decides. That’s the pattern that enterprises trust. That’s the pattern that actually solves problems worth solving.

The Three Preconditions for Agentic Success

The panelists from ServiceNow, Kinaxis, Langfuse, and CIBC laid out a framework—three things you need in place before agents can work at scale:

1. Data in Order
Not perfect. In order. You need to be able to feed agents clean, relevant information and trust that they’re making decisions on good inputs.

2. Modeling That Drives Right Behaviors
This is where domain knowledge becomes infrastructure. Kinaxis talked about embedding decades of supply chain expertise into an ontology—a semantic layer that constrains how agents behave. Without this, agents will generate “nonsensical” behaviors that look logical from a pure math perspective but violate real-world constraints. The heavy lift isn’t the model; it’s translating human experience and domain rules into a system the agent can navigate.

3. Transformed Ways of Working
Deploying agents isn’t a technology upgrade; it’s an organizational transformation. Humans and agents need to work together in new patterns. You need to rethink approvals, accountability, feedback loops, and decision rights.

Measuring an Agent’s Readiness

One of the most honest moments came when panelists discussed how to know if an agent is ready for production. The answer: You need the capacity to evaluate, measure, and trace its performance.

This means:

• Building synthetic test cases to validate behavior
  
• Extracting real-world datasets to test against
  
• Creating benchmarking test suites that persist over time
  
• Using internal feedback to label agent outputs and close learning loops

It’s not “set and forget.” A panelist from Langfuse emphasized that agents can make mistakes once and learn, but only if you close the loop deliberately. You have to test to ensure they don’t repeat the same mistakes. This requires organizational discipline that many companies don’t yet have.

This is also where ROI gets proven. Enterprise adoption doesn’t happen on faith. It happens when you can show that the time saved translates to measurable business impact. In manufacturing, that’s labor hours freed up to work on preventive maintenance instead of reactive firefighting. In construction, it’s project managers moving from manual reconciliation to strategic problem-solving. The enterprises that move fastest are the ones where you can quantify: “This agent saves your team 8 hours per week on schedule analysis. At fully loaded cost, that’s $X annually. Here’s what you do with 400 hours of reclaimed productivity.” That math needs to be airtight. And it needs to be tested and proven before you ask them to rely on it enterprise-wide. The companies that scale are meticulous about this proof.

The Semantic Layer as Secret Weapon

One of the most insightful comments came from Kinaxis: “Semantic systems exist in inhuman heads. Creating that layer in tech is where the heavy lift is.”

This is the insight that separates companies scaling agents from those stuck in pilot purgatory. And it’s the statement that made me sit back, because I’ve lived this exact challenge for years.

Building agents isn’t about writing clever prompts. It’s about translating deep domain knowledge—the kind that lives in experienced people’s heads—into explicit, bounded systems. In construction, that knowledge includes:

• Site constraints (access points, weather windows, equipment limitations)
  
• Historical patterns (how long certain tasks actually take versus what the schedule says)
  
• Regulatory requirements (safety protocols, certification hold-ups, inspection dependencies)
  
• The ripple effects (how a concrete delay cascades to framing, which delays electrical, which delays finish work)

This is why so many AI projects fail in regulated industries. Companies try to deploy agents without understanding the domain deeply enough to codify the rules. The agent makes technically correct decisions that are operationally nonsensical. A schedule optimization that violates union rules. A procurement decision that breaks compliance.

This is where your product roadmap becomes your marketing roadmap. The enterprises that adopt are the ones who see that you’ve embedded decades of industry knowledge into the system. They see themselves reflected in the constraints and logic. That validation—“they understand how we actually work”—is what moves enterprises from skepticism to championship.

In my work leading marketing for AI companies in these spaces, the ones who scaled fastest weren’t selling “we have better AI.” They were selling “we spent 18 months with your peers, your competitors, and domain experts understanding how this actually works, and we built that into every decision the system makes.” That story resonates because it’s credible. It shows you’ve done the hard work of translation.

Accountability, Governance, and the Human

CIBC brought a critical perspective: “Accountability falls to the human. No matter the tech.”

This means:

• Humans remain responsible for agent outputs
  
• You need access controls and performance management frameworks for agents (approval limits, decision categories) just like you do for humans
  
• The database integration work—connecting agents to legacy systems—is often harder than the AI part

And here’s where it got real: the panelists acknowledged that agents can make mistakes that look like successes. A fraudulent transaction approved by an agent looks different from a fraudulent human approval, but the damage is the same. This is why monitoring, flag systems, and guardrails specific to your application are table stakes.

This is non-negotiable in regulated industries, and it’s also where trust is built or broken. When I worked with enterprise AI companies in construction and manufacturing, the conversations weren’t about capability—they were about governance. “How do I ensure compliance?” “What happens if something goes wrong?” “How do I audit the decisions?” The companies that won were transparent about these questions. They showed audit trails. They built dashboards. They admitted limitations. They designed governance frameworks before deployment, not after. That’s what gave enterprises permission to trust. And that’s what you need to communicate in your marketing—not “our AI is better,” but “our AI is governed in ways that align with how you already work.”

The Emerging Problem: Agent-to-Agent Interactions

One panelist raised a question that hints at the frontier: What happens when agents interact with other agents? We’re early enough that we don’t have good answers, but it’s clear that emergent behaviors will arise. Organizations need to be deliberate about the culture they want to enable—how humans interact with agents, and how agents interact with each other.

This is territory we haven’t collectively figured out yet. But from a marketing standpoint, it’s where positioning matters. Early adopters in regulated industries will be cautious here—rightfully so. They’ll want to work with vendors who’ve thought deeply about this problem, who are transparent about what they don’t know, and who build in guardrails and observability before scaling agent-to-agent workflows. That’s a positioning opportunity for vendors who are willing to be honest about the frontier.

The Through-Line: Intentionality at Scale

Both talks, from different angles, pointed to the same underlying challenge: moving AI from experimentation to enterprise scale requires deliberate, foundational work across technology, data, talent, and culture.

The companies positioning themselves to win aren’t waiting for perfect conditions. They’re:

• Investing in talent and empowerment — hiring people with authority, giving them context, and letting them experiment with risk tolerance
  
• Building foundational infrastructure — data pipelines, semantic layers, governance frameworks
  
• Transforming operations — not just adding AI on top of existing processes, but rethinking how humans and machines work together
  
• Closing feedback loops — measuring obsessively, benchmarking continuously, and treating deployment as the beginning of learning, not the end
  
• Identifying ideal customers — finding enterprises with champion executives, clear business problems, and appetite for transformation, then making them the centerpiece of go-to-market strategy

But here’s what I want to emphasize, because it’s critical: identifying and marketing to the right customers is where the magic happens.

Not all enterprises are ready to adopt AI. And that’s okay—that’s actually good news if you know how to spot the ones who are.

The enterprises I worked with that scaled fastest in construction, manufacturing, and regulated industries shared specific traits:

• They had a clear, quantifiable problem that was costing them real money
  
• They had leadership willing to admit the problem was bigger than internal solutions could fix
  
• They had domain expertise they could share to help you build the right solution
  
• They had patience for a transformation journey, not appetite for a quick fix
  
• They had a champion executive—someone who’d fought for modernization before, who could rally their peers, and who would evangelize the change

Those customers don’t just adopt your product. They become partners in your roadmap. They speak at conferences about you. They serve as references for peers. They turn into evangelists because they’ve invested in the journey and they’ve seen the results.

This is where marketing meets product strategy. The companies winning aren’t spreading themselves thin trying to convince everyone. They’re ruthlessly focused on finding and prioritizing those ideal customers. They’re investing marketing dollars into reaching the ones with champion executives and clear pain points. They’re measuring go-to-market efficiency by the quality of customer relationships, not just the number of logos.

What’s Next

As I headed into Day 2 of ALL IN, I left Day 1 with a sharper sense of where the real work lies. It’s not in the labs. It’s in the boardrooms where CEOs and CFOs align on AI strategy. It’s in the data warehouses being rebuilt. It’s in the teams redefining what oversight looks like when machines have agency.

And it’s in the marketing strategies that recognize that selling AI to enterprise buyers, especially in regulated industries, requires patience, domain knowledge, and an obsession with solving problems worth solving. It requires understanding that your ideal customer isn’t always the biggest company. It’s the one with the champion, the clear business case, and the willingness to transform.

ALL IN reminded us all in attendance that this conversation is happening in Canada, with Canadian founders and leaders building real, scaled applications. That’s worth paying attention to.

Let’s Talk

I work with Founders and Startup Executives to market your product or service to grow and reach your next milestone or raise. I unlock new strategies and tactics to drive growth that are rooted in my years of hard-won lessons: leading marketing at AI product companies selling to enterprises in construction, manufacturing, and other regulated industries. I’ve learned how to identify the barriers to adoption, how to translate domain expertise into compelling product positioning, how to speak credibly to CFOs and CEOs, and how to spot and prioritize the customers ready to champion AI transformation.

If you’re an AI company selling to enterprise buyers and you need someone who speaks both the language of your customers and the language of product and go-to-market strategy, that’s exactly the problem I solve. I understand the constraints, the governance requirements, the ROI questions, and I understand how to position and market to customers ready to move.

What from ALL IN resonated with you? DM me on LinkedIn or drop a comment.

Let’s Talk