From Pilots to P&L — The Five AI Conversations That Matter Right Now

Koko: Eighty-eight percent of organizations are using AI in at least one function. Only thirty-nine percent can point to any measurable impact on earnings. That gap — roughly half the business world using a technology that isn't showing up in profit and loss — is one of the most consequential numbers in corporate strategy right now.

Sam: And the reaction most executives have when they hear that is — wait, which half am I in?

Koko: Exactly. And the uncomfortable answer from the research is: probably the first half. Using AI, not necessarily profiting from it.

Sam: Which sounds damning, but there's a more interesting version of that finding. The problem isn't the technology. It's the management.

Koko: That is the whole argument. And it is what the piece we're digging into today is built around. It's called From Pilots to P&L — The Five AI Conversations That Matter Right Now, and the thesis is essentially: the debate has moved. It is no longer about whether to adopt AI. It is about whether you can make it accountable.

Sam: Which is a much harder problem. Adoption you can declare. Accountability you have to earn.

Koko: So the piece lays out five conversations — five topics where, if you walk into a boardroom with a crisp, defensible point of view, you win the room. And the thread connecting all five is that same gap: massive adoption, thin impact.

Sam: And the spending is accelerating into the gap, not away from it. BCG's AI Radar for twenty-twenty-six found companies plan to roughly double AI spend this year — to about one point seven percent of revenue. That is a meaningful number.

Koko: And ninety-four percent of those companies intend to keep investing even without near-term returns.

Sam: Which is either admirable conviction or expensive faith, depending on your vantage point.

Koko: The piece is pretty clear on which. Half of CEOs believe their job depends on getting AI right. So there is real personal urgency here. But urgency without accountability is just fast spending.

Sam: And that is what the pilot-failure numbers really show. MIT's research put the rate of AI initiatives with no measurable P&L impact near ninety-five percent — though the piece is honest that the methodology is contested. McKinsey's thirty-nine percent is the more conservative, friendlier-faced version of the same story.

Koko: Two very different numbers pointing in exactly the same direction.

Sam: Which is the direction of: the pilot era is over. The reckoning has started.

Koko: Let's walk through the five. The first one is the ROI reckoning, and the take is pointed: the AI ROI problem is not a model problem. It is a management-accounting problem. Which makes it a CFO problem.

Sam: That reframe alone is worth the price of admission. Most organizations treat AI investment like R&D — you fund it, you wait, you hope. The piece is saying treat it like a capital program with a benefits ledger, workflow-level unit economics, and FP&A attestation.

Koko: If a benefit can't survive an FP&A review, it's not a benefit. That's the line. And I think that will land very differently for a CFO than for a Chief AI Officer.

Sam: Because for a CFO it's not a new idea — it's just a familiar discipline applied somewhere it hasn't been applied yet.

Koko: Second conversation: agents. McKinsey finds twenty-three percent of organizations are scaling agentic AI somewhere, but no more than ten percent are scaling it in any single function. And here is the number that really lands — PwC found seventy-nine percent of executives say agents are being adopted, but only thirty-four percent use agents in finance at all.

Sam: And only forty-two percent are even redesigning processes to accommodate them. So agents are in the building, but the building hasn't changed.

Koko: Deloitte adds that eighty-four percent of companies have not redesigned a single job around AI. Not one.

Sam: Agents dropped into unchanged workflows produce demos, not leverage. That's the piece's line, and it's pretty good.

Koko: Third — and this one genuinely surprised me — the disruption to professional services pricing. The billable hour is melting. When sixty hours of analysis becomes six, time stops being a proxy for value.

Sam: And the piece gives you the receipts. McKinsey reportedly gets more than thirty percent of global fees from outcome-tied pricing. BCG has said AI-related work is moving from roughly twenty percent of twenty-twenty-four revenue toward forty percent. EY is openly describing a 'service-as-software' future.

Koko: And then the hyperscalers entered the arena. On July second, Microsoft launched what it calls Frontier Company — two point five billion dollars, six thousand embedded experts, explicitly an outcome-driven engineering organization.

Sam: Two days after AWS stood up a one-billion-dollar embedded-engineer unit. The direction is unmistakable, as the piece says. The firms that convert expertise into reusable, governed, priceable assets — ontologies, diagnostics, agent suites — compound. The firms still selling hours hand their productivity gains to the client.

Koko: That is a brutal and probably accurate summary.

Koko: The fourth and fifth conversations are where the piece gets into the territory that I think is most underappreciated right now. Controls and compliance. And this is where I want to push back a little, because I think the 'human in the loop' argument is the most provocative thing in the piece.

Sam: Go on.

Koko: The piece says human-in-the-loop is becoming the industry's favorite checkbox. A human who lacks the authority, context, time, and tooling to actually detect and stop an agent is not a control — they are an audit finding waiting to happen.

Sam: And regulators are now saying the same thing out loud. Bank of England Deputy Governor Sarah Breeden, at the ECB's Sintra forum on June thirtieth, said existing frameworks were not built to contemplate autonomous agents, and that relying on a human in the loop for every agent action is 'unlikely to be realistic.'

Koko: She even floated market-wide circuit breakers for AI-driven trading. That is a significant statement from a senior central banker.

Sam: And the FSB's June consultation is recommending treating agents as synthetic employees — with boundaries, approval gates, audit trails. This is not theoretical governance. This is incoming infrastructure.

Koko: McKinsey finds high performers are nearly three times more likely to have defined human-in-the-loop validation — sixty-five percent versus twenty-three percent. So the performance gap maps directly to the control gap.

Sam: Here's where I'd push slightly differently. The piece frames this as mostly a financial-controls question — ICFR thinking applied to agents. And that's right. But I'd argue it's also a talent question that hasn't been named yet. If agents are a workforce, who manages them? Do you have people who actually know how to supervise a non-human employee?

Koko: The piece doesn't go there explicitly, but that absence is itself interesting. Eighty-four percent of companies haven't redesigned a single job — so the answer is probably no, they don't have those people yet.

Sam: And then you layer on the EU. Which brings us to conversation five — and the piece's most useful reframe of a news story.

Koko: The Digital Omnibus. The EU delayed high-risk AI obligations — the Annex III stuff — from August of this year to December twenty-twenty-seven. And the industry read it as relief.

Sam: The piece says that's exactly backwards. A deferral is a scoping window, not a pause. General-purpose-AI obligations have applied since August twenty-twenty-five. Transparency obligations still land August second twenty-twenty-six. Generative-content watermarking is due December second twenty-twenty-six.

Koko: So the deadline moved. The obligation didn't.

Sam: And the companies that treat it as a pause will arrive at December twenty-twenty-seven with an unbuilt inventory, unclassified systems, and a queue at the notified bodies. That is not a fun place to be.

Koko: Governance embedded in delivery scales. Governance parked in a policy PDF does not. That's the formulation in the piece and it's one of those lines that sounds obvious until you realize almost nobody is acting on it.

Sam: So what does a smart executive actually do with this? Because the piece covers a lot of ground and it would be easy to nod along and then go back to the innovation dashboard.

Koko: Right. So here's my honest read. The most actionable thing in the piece is the CFO reframe. If you are a CEO or a board member, the question to ask right now is not 'are we using AI' — you are, almost certainly. The question is: does your finance function have a benefits ledger for AI? Workflow-level unit economics? If the answer is no, you have an accountability gap, and you should name it that.

Sam: And on agents — stop asking 'have we deployed agents' and start asking 'have we redesigned any work around them.' Those are very different questions with very different answers.

Koko: For professional services executives — law firms, consulting, advisory — the pricing conversation is urgent. The piece is not saying your model collapses tomorrow. It is saying Microsoft and AWS just made explicit what was implicit. The clock is running.

Sam: And on the regulatory side, even if you're not primarily EU-facing — the control logic the FSB and Bank of England are articulating will migrate. Treat agents as synthetic employees is not a fringe position anymore. Build toward that now while you have runway.

Koko: The piece's summary is actually quite clean: measure AI like capital, manage agents like labor, price outcomes instead of hours, make controls real instead of ceremonial, and use the regulatory runway to build instead of wait.

Sam: Five verbs, no buzzwords. Genuinely hard to argue with any of them.

Koko: The only one I'd add a caveat to is the pricing point — because repricing your advisory model when your clients haven't asked you to takes real nerve. It's the right strategic move, but it requires a CFO who's willing to shrink the top line before the market shrinks it for you.

Sam: Which is basically the whole history of technology disruption in three sentences.

Koko: The number that keeps coming back to me from this whole piece is not the adoption rate or the failure rate. It's that half of CEOs believe their job depends on getting AI right — and they are still doubling spend without the accounting infrastructure to know if it's working.

Sam: That is a high-stakes bet on a black box. And the piece's argument is that the black box is optional. You can open it. It's a management problem, not a mystery.

Koko: Which is either reassuring or more alarming, depending on the quality of your management.

Sam: The question I'd leave listeners with is this: your company almost certainly has an AI adoption story. Does it have an AI accountability story? Because those are not the same story, and only one of them matters to the board in twenty-twenty-six.

Koko: That is the right question. From Pilots to P&L — The Five AI Conversations That Matter Right Now is the full piece, and every number in it is source-verified — McKinsey, BCG, PwC, Deloitte, the FSB, the Bank of England. You can find it on Koko Knows. Read the whole thing. Bring the five conversations to your next leadership offsite. See who in the room has a real answer.

Sam: And if nobody does — that's your answer.