ChatGPT
Perplexity
AI arrives looking like a technology purchase. It gets funded that way, staffed that way, and reported on that way. The return follows a different logic.
Microsoft’s Work Trend Index measured what predicts who reports real AI impact. Organizational factors like culture, workflow design, and talent development weighed more than twice as heavily as individual skill and mindset—67% against 32%.
None of that shows up on a purchase order, which is why it rarely gets a budget line or a named owner.
This piece is about that two-thirds. The part of AI’s return no purchase order covers, and where compounding returns come from once the tools are in.
Three of the four foundations behind AI returns sit outside the tech team
Most AI programs are rolled out as technology deployments with a training plan attached. The results keep pointing elsewhere.
PwC’s Global CEO Survey asked 4,454 CEOs what they were getting from AI. 56% reported no revenue or cost benefit at all. The one in eight who captured both did not have better models.
They had a responsible AI framework, a defined roadmap, a technology environment built for enterprise-wide integration, and a culture that enabled adoption.
Three of those four foundations sit with the business. Only the last belongs to your technology team.
From the field.
When Moderna set out to make AI change how work got done, it merged human resources and IT into a single function, put a former HR chief in charge of both, and began sorting the work itself into what people should do and what machines should do. The reorganization was the strategy. The tool ran inside it.
Moderna is far larger than a mid-market company, but deciding who owns the shape of work is a choice a 500-person firm makes faster than a 40,000-person one.
Do this next.
Before the next license is approved, take stock of what actually moves AI value beyond the tool itself: integration-ready systems, a named governance owner, and an adoption plan with a business owner behind it. Then check which of those has a real name attached today. The gaps you find are usually the ones no budget line covers.
Stay updated with Simform’s weekly insights.
AI budgets are outpacing the discipline behind them
AI budgets rise because the case for AI feels clear. What gets built inside the program often doesn’t keep up.
Grant Thornton’s 2026 survey found that 83% of finance leaders reported increases in AI budgets heading into 2026. The same survey found that 78% of leaders could not confidently pass an independent AI governance audit within 90 days.
Read those two figures together, and the shape of the problem gets specific. The funding to run AI has been approved. The discipline to know whether it is working is not yet in place.
In most organizations, the gap opens because approving a budget is a single decision made once, while building measurement discipline is an operating change that must be sustained quarter after quarter. No one has been asked to own the second.
Each new deployment lands on a foundation that isn’t yet tracking what worked. The cost side of AI compounds long before the return side does. For a CFO watching next year’s forecast, the question worth asking is which pieces of the program have a review cadence attached, and which ones will still be running unmeasured this time next year.
Where AI governance ownership sits shapes AI’s financial impact
Grant Thornton’s AI Impact Survey, with a dedicated segment of mid-market companies at $100M to $1B in revenue, found governance to be the single most-cited reason AI programs underperform. It ranked ahead of insufficient training and ahead of weak data readiness.
That ranking is worth sitting with. Training and data are the two things most AI programs pour money and attention into first. Governance is what mid-market leaders report is costing them value.
The reason has less to do with the policies themselves than with where the authority to enforce them lives. When a technical team leads governance, the decisions it produces remain technical, focused on model access rules and deployment guardrails.
When it sits at the executive level, it starts producing the harder decisions that determine returns. Which AI initiatives get funded next quarter, and which ones get killed when the pilot doesn’t clear the bar. Those are business decisions, and they only stick when they carry executive weight.
What can you do?
The move for this quarter is to name one senior leader responsible for AI outcomes, ideally from the business side. Set risk criteria centrally, let trained reviewers handle case-by-case calls, and ensure every agent in production has a business owner who remains accountable for its performance after it goes live.
AI returns compound when workflows travel between teams
There is a difference between rolling out AI and absorbing it, and the durable advantage lives in the gap between them.
Every agent and workflow you run throws off signal. What worked, what failed, where a human had to step in. In most organizations, that signal stays inside the team that produced it, and the next team rediscovers the same lessons from scratch.
McKinsey’s 2025 analysis of 25 adoption practices found workflow redesign to be the single practice most strongly correlated with bottom-line impact from generative AI. The redesign is not the workflow itself. It is the mechanism that lets one team’s working setup become the default the next team inherits.
What can you do?
For your highest-volume AI workflow, three ownership questions decide whether that mechanism exists. Who reviews how the agents are performing, who can change the workflows they run, and how a win on one team becomes the version the next starts from.
Answer those three, and a program of scattered pilots turns into something a competitor cannot replicate by buying the same tools.
Mid-market structure is the AI advantage enterprises pay to recreate
Mid-market companies have the structural advantage enterprises spend fortunes trying to buy back. Fewer layers between a decision and the floor, faster changes of direction, leaders who still talk directly to the people doing the work.
Grant Thornton’s mid-market cut found that firms with fully integrated AI are close to four times more likely to report revenue growth than those still running pilots, 58% against 15%.
What separates those two groups is not budget size. It is whether leadership used their proximity to reshape how the work runs, or spent the budget without touching the shape of the work underneath it.
The question worth carrying into your next executive meeting is narrow. Of the levers this piece covered, which ones have a name attached today, and which are sitting unclaimed in the space between the COO and the CFO?
Assigning ownership and rebuilding how the work runs around it is the engagement. Simform’s Enterprise AI adoption and change management advisory designs the governance, ownership, workflow redesign, and adoption plan that determines whether an approved AI program compounds or stalls around month three.
