Unblocking AI Transformation: What I’m Seeing Inside Enterprises Right Now
by
Sheetal Jaitly
Over the past year, I’ve spent a lot of time inside large enterprises working on AI initiatives—across different industries, different levels of maturity, and very different starting points.
And despite all that variation, I keep seeing the same pattern.
Most organizations are not struggling to start with AI. They’ve already done that.
They’ve run pilots. They’ve rolled out tools. Teams are experimenting. There are real, tangible wins—faster workflows, reduced manual effort, better access to information.
But those wins don’t scale.
AI stays stuck in pockets instead of becoming part of how the business actually operates.
What’s becoming clear to me is that this isn’t a technology problem. It’s an execution and operating model problem.
Pilots Are Easy. Production Is Hard.
One of the biggest gaps I see is the distance between a successful pilot and a production-ready system.
Pilots are controlled. They’re scoped. They’re often built by a small, motivated team working around constraints.
Production is different.
Now you’re dealing with real workflows, real risk, real accountability. You need reliability, governance, integration, and clarity on who owns what.
A lot of organizations underestimate that jump.
So they keep running more pilots—hoping the next one will somehow “unlock” scale.
In my experience, it rarely does.
You Can’t Layer AI Onto Broken Workflows
Another consistent pattern: AI is often applied to processes that were already inefficient.
That usually leads to incremental gains, but not transformation.
The more effective approach—and what we’ve been leaning into at TribalScale—is to step back and redesign the workflow entirely.
If you assume AI agents and humans are collaborating from day one, the process often looks completely different.
You start asking different questions:
What should AI own end-to-end?
Where does human judgment actually matter?
How do you design clean handoffs instead of patching gaps?
That shift—from optimization to redesign—is where the real leverage is.
Speed Changes the Quality of Decisions
One thing I’ve become increasingly convinced of: speed is not just about moving faster—it actually improves outcomes.
At TribalScale, we work in small, senior pods—typically strategy, product, and engineering—embedded directly with business units.
Instead of long discovery phases, we run tight, focused sprints.
The goal is to go from idea to something real, usable, and production-ready in a matter of weeks.
Working this way forces prioritization. It surfaces constraints early. And it keeps everyone aligned on outcomes instead of abstractions.
More importantly, it builds momentum inside the organization. People see something working in their environment—not in a demo, but in their day-to-day workflow.
That changes the conversation entirely.
The Real Bottleneck: Organizational Knowledge
Another thing that doesn’t get talked about enough is how much critical knowledge inside enterprises is still implicit.
It lives in people’s heads—in how they make decisions, handle exceptions, or navigate edge cases.
When we start building AI systems with business teams, this becomes immediately obvious.
You can’t build something reliable without that context.
So a big part of the work ends up being extracting and structuring that knowledge—pairing operators with builders to make it explicit.
Interestingly, when this is framed as capturing expertise and shaping the future system, people are much more engaged than expected.
It stops feeling like replacement and starts feeling like contribution.
AI Needs to Be Managed Like a Workforce
Another shift I’m seeing: thinking about AI as “just a tool” breaks down pretty quickly at scale.
AI agents start to behave more like a workforce.
They take on tasks, make decisions, and produce outputs that need to be reviewed and improved over time.
That raises new questions:
Who is responsible for their performance?
How do you measure quality?
What are the boundaries of what they’re allowed to do?
The organizations that are getting traction are the ones starting to treat AI this way—defining roles, permissions, and accountability structures, not just deploying models.
Role Redesign Is the Missing Piece
Even when AI works technically, value doesn’t always follow.
Why?
Because roles haven’t changed.
If people save time but continue doing the same type of work, the impact is limited.
The real opportunity is in shifting how time is spent—toward higher-value activities like decision-making, system design, and oversight.
But that doesn’t happen automatically. It needs to be designed and reinforced.
What’s Actually Working
If I step back and look at what’s consistently working across organizations, a few things stand out:
Starting with real business workflows, not abstract use cases
Working directly with business units, not in isolation
Moving quickly from idea to production—not getting stuck in pilots
Using small, experienced teams that can both decide and build
Treating AI as part of the operating model, not an add-on
That’s the model we’ve been applying at TribalScale, and it’s what’s allowed us to activate production-ready AI systems in weeks.
Not perfect systems—but real ones. Systems that are live, improving, and creating value.
Closing Thought
AI transformation isn’t blocked by a lack of tools or ideas.
It’s blocked by how organizations approach change.
The longer companies stay in exploration mode, the harder it becomes to shift.
At some point, the only way forward is to start building—inside the business, with real constraints, and with the expectation that what you create will actually run.
That’s where things start to compound.