Creating AI Value: What’s In It For You?

The slide deck is good, the vendor has done their part – demo lands, case studies check out, timeline is realistic enough to be believable – and the internal sponsor walks through it like someone who has spent a few weekends rehearsing. There’s an AI initiative on the table, real budget attached, and the room is mostly nodding along. Then the CFO asks the obvious question: “Okay, but what’s the value?”

The room goes quiet, not because nobody has an answer, but because everybody has one, and that’s actually the problem. The sponsor has a number, the IT director has a different number, the head of operations has yet a third take, and a few of those numbers aren’t really answers at all – they’re hopes, dressed up in spreadsheet clothes, with a productivity assumption baked in somewhere that nobody wants to defend out loud (if AI makes my sales reps just five percent more productive …).

What’s in it for me? What’s in it for the company? You’d think after forty years of pitching technology investments to skeptical finance leaders, the industry would have a cleaner way to answer that question, but we don’t, and AI hasn’t made it easier – it has made it harder. The hype-to-results gap is wider than anything we’ve seen since the dot-com era, ROI is squishier because the work itself is changing (not just getting faster), and the soft-dollar hand-wave that limped through every ERP business case from the 1990s won’t survive contact with a halfway competent CFO this time around. Or at least, it shouldn’t.

So let’s slow down and try to answer the AI project value question honestly. Because the value of an AI initiative isn’t one number. It’s three different conversations with three different stakeholders, and the honest answer involves all three of them.

The question every AI project has to answer

Every project, AI or otherwise, has to answer the value question, and the standard playbook has always offered three flavors of answer: revenue growth, cost reduction, or risk management. I wrote about this back when ERP was still the big technology bet on the table, and the math hasn’t really changed – if your project doesn’t grow the top line, reduce a cost somewhere, or manage a real risk to the business, you’re going to have trouble justifying it. Sometimes a project hits two of the three, occasionally all three, but at least one of these levers needs to show up in the value story or the project doesn’t move.

AI doesn’t change those three categories. What AI changes is the honesty test you have to apply when you fill them in. Revenue growth used to mean “we’ll sell more widgets because our sales team has better information,” which was fine as long as you could connect the system to a measurable change in close rates or order size. Cost reduction used to mean “we’ll need fewer people doing this task” or “we’ll consume less of this raw material,” both of which a finance team can audit. Risk management used to mean “we’re staying out of jail and off the front page,” which is easier to defend than to quantify but still defensible. Each lever had a known way of being lied to, and a known way of being honest.

AI invites a new kind of fuzz. The productivity story is back, dressed up in fresh language: “AI will give every knowledge worker a 30% lift” sounds modern and confident, but it’s the same soft-dollar hand-wave that justified two decades of mediocre CRM rollouts, just with a more impressive demo behind it. The cost-reduction story has its own twist: vendors will tell you that “five bots can replace eight people,” and sometimes that’s true, but more often the bots replace pieces of work and the eight people stay busy with the work that remains. And risk management has flipped on its head – for the first time in a long time, the biggest risk isn’t doing something. It’s not doing something while a competitor does it well.

So the question is still the same question: what kind of value, exactly, and how will we know? But the answer needs more rigor than it used to, because the failure mode for AI projects isn’t a clean miss against a clean number. It’s a slow drift into “well, it’s helping somewhere, we think,” which is a terrible place to end up after spending real money. The honest practitioner has to push past the three categories and ask the next question: value for whom?

Why AI project value matters more now than ever

The temptation is to wave all this off as the same value conversation we’ve had for every wave of technology. ERP, the Internet, mobile, cloud, IoT – same playbook, same hand-wringing, same eventual settling in. That would be fair, except the numbers this time around are starting to look genuinely different.

BCG has been tracking what they call “future-built” companies, the small group running AI well, versus everyone else¹BCG, “The Widening AI Value Gap: Build for the Future 2025”. The story isn’t that AI is magic; it’s that the companies running it well have figured out how to convert capability into compounding advantage.. The headline numbers are striking: leaders are seeing 5x the revenue increases and 3x the cost reductions that everyone else gets from AI. Only 5% of companies qualify as future-built. Another 35% are scaling and starting to see gains. The remaining 60% are reporting minimal returns and don’t yet have the capabilities to scale. McKinsey’s “Rewired and running ahead” research lands in roughly the same place on the gap, with TSR deltas of 2x to 6x depending on industry. Those aren’t gaps that close on their own.

Here’s why that matters, practically. Your CFO has almost certainly seen these numbers, or numbers very much like them. Big-firm research filters into board decks and earnings calls, and “what are we doing about AI” is now a standard question. Saying “we’re being careful” doesn’t cut it anymore. Being careful is what your slower competitors are saying too, and the gap is opening anyway.

Joel Trammell, who spent years as a CEO and now writes about how boards actually think, makes a point worth sitting with in a recent piece on AI and the boardroom. Boards are increasingly looking at AI capability as a leading indicator of future earnings, not just an operational improvement. The market is pricing in the gap before it fully shows up in the financials. Which means the cost of being late isn’t just the project you didn’t do. It’s the multiple compression you wear while you catch up. That’s a different kind of risk than IT departments are used to managing.

None of this means panic-buying an AI strategy. The same research that shows the leader-laggard gap also shows what the leaders do differently, and it’s almost never the technology choice that separates them. It’s the discipline of connecting AI investments to specific value outcomes. Revenue growth in a definable segment. Cost reduction in a measurable process. Risk management against a real exposure. Then sticking with the value conversation long enough to actually deliver. Which brings us back to the harder question, the one nobody really wants to open up in the middle of a budget cycle: value for whom?

Great Stock or Great Company?

Here’s a question that doesn’t get asked often enough, and probably should: are you building a Great Stock, or a Great Company? It’s worth pausing on, because the answer most leaders give in public is “both, of course,” and the answer most companies live out in practice is “one, depending on the quarter.”

A Great Stock is the raw, capitalist form of value. It’s measured in total shareholder return, driven by earnings growth, capital efficiency, and the multiple the market gives you. The audience is finance – analysts, institutional investors, the board’s compensation committee. The time horizon, depending on who you ask, is somewhere between next quarter and the next three years. The metrics are hard, the conversations are crisp, and there’s a long history of well-understood ways to move the needle. AI fits into this conversation as a productivity story, a cost story, and increasingly as a multiple story (the leaders are pulling away, remember?).

A Great Company is the other half of the picture. The employees know where the business is going, they trust their managers, and they have the tools and processes they need to do their work well. The customers are listened to, their problems get solved, and the products keep getting better in ways that matter to them. The metrics are softer – engagement scores, NPS, retention, the things HR and customer success people present to a polite-but-skeptical executive team. But the engine that makes Great Stock sustainable over any meaningful time horizon is, frankly, Great Company. Engaged employees doing terrific work for happy customers – that’s what generates the earnings the shareholders are getting excited about in the first place.

The tension between the two isn’t theoretical. A short-term focus on quarterly earnings can drive the stock value, while quietly making it tough on the people and the customers who actually create that value over time. Cut training budgets, hold off on hiring, defer the unsexy infrastructure work, sweat the workforce a little harder – the math works for a while, and then it doesn’t. You can only cut so far. And the moment a competitor with a better-engaged team and stronger customer relationships starts pulling away, the multiple compression hits anyway, and now you’ve got both problems at once.

On the other hand, we are running a business here, and there are times when the company needs the employee and times when the employee needs the company. Some decisions look harsh from the inside. Pivots away from favored customers, popular projects, comfortable markets – these create change that teams won’t entirely agree with, and shouldn’t have to. The job of leadership isn’t to pretend the tension doesn’t exist. It’s to navigate it with open communication, clarity of purpose, and a focus on maintaining the balance over time.

AI sharpens this tension in a way that previous technology waves didn’t. The cost-reduction story has never been more vivid (vendors will happily tell you which roles you can eliminate), the productivity story has never been more seductive (every demo shows a knowledge worker getting 10x more done), and the speed of change has never been faster. Which means the leaders who default to the Great Stock conversation will be the ones cutting first and thinking second. And the leaders who default to the Great Company conversation will be the ones explaining, six months from now, why their AI program is “still in pilot.” Neither of those is the right answer. The right answer is harder, and it starts with naming the three stakeholders honestly.

The three stakeholders, briefly

Three stakeholders, three different value conversations. The brief versions look like this, and each one will get its own article in the series ahead.

Shareholders care about total shareholder return, and TSR breaks down into three levers. Earnings growth – the day-to-day business of creating and selling at a fair profit, where AI shows up in operations efficiency and sales effectiveness. Capital efficiency – what the business does with the cash it has, where AI shows up in better forecasting, leaner inventory, and smarter capital allocation. And multiple expansion – the gap between what the company is worth on paper and what the market is willing to pay for it, where AI capability is increasingly being priced in as a leading indicator (Trammell’s point from earlier in this article). Of those three levers, the multiple expansion conversation is the one most leaders are least prepared for, and the one most likely to surprise them in the next 18 months.

Employees care about engagement, which sounds soft until you look at the math. Bain’s research on “putting people first” shows companies that lead with their teams deliver about 2.3x the TSR of companies that don’t. That’s a hard number wrapped around a soft idea, and it’s the bridge that connects Great Company back to Great Stock. The engagement question itself breaks down into three honest sub-questions: do I know where we’re going, do I get the support I need from my manager, and do I have the tools and processes to do my job well? AI changes the answer to all three, sometimes for the better and sometimes for the worse. The work itself is changing – not just the speed of the work, but the substance of it – and the employees who used to be very good at the old work are not automatically going to be very good at the new work. Engaged teams adapt. Disengaged ones quietly check out, and then suddenly check out for real.

Customers care about getting their needs met, which AI actually does help with in ways previous technology waves couldn’t. Satya Nadella has been making the Customer Obsession argument for years, and it has only gotten sharper as AI capabilities have matured. There are two halves to the customer value conversation. The connection side – really listening, anticipating needs, reducing friction, treating customer service as a feature rather than a cost center. And the product side – building information and intelligence into the products themselves, so they get better with use, predict failures before they happen, and create value the customer didn’t know to ask for. AI is uniquely well-suited to both halves. The companies that get this right won’t just have happier customers. They’ll have customers who would have a hard time imagining going back.

Three stakeholders, three conversations, three sets of metrics. None of them are optional. And the honest leadership team has to be prepared to have all three conversations, not just the one that fits this quarter’s narrative.

The leadership conversation, not the spreadsheet

The mistake most leadership teams make, when they finally sit down to have the three-stakeholder conversation, is to try to settle it with a spreadsheet. Take the AI initiative on the table, build a ROI model, factor in some employee impact, sprinkle in a customer experience score, and roll it all up into a single number that supposedly answers the value question. Then someone changes one assumption and the number swings 40%, and everyone goes quiet.

The spreadsheet isn’t the wrong tool, exactly. It’s just the wrong conversation. A three-stakeholder value analysis isn’t an optimization problem with a single answer. It’s a leadership conversation with three honest answers that sometimes contradict each other, and the job of the leadership team is to navigate those contradictions in plain language – not to make them disappear inside a formula.

Here’s what that conversation actually sounds like, when it’s working. Someone names the shareholder case: “This AI initiative drives revenue in our top-three customer segment, and we think it adds 80 basis points to operating margin within 18 months.” Someone names the employee case: “It changes what our service team does day-to-day, and we’ll need to retrain about 40 people. A few of those roles will shift substantially, and we owe those folks a clear conversation now, not six months from now.” Someone names the customer case: “It cuts our average response time by 60% and lets us catch issues before customers notice them, which our biggest accounts have been asking for.” And then someone, ideally the person running the meeting, asks the question that ties it together: “Are all three of those true at the same time, or are we trading one against the others?”

That last question is the one that earns its keep. If all three are true, you have a strong project, and you should move fast. If you’re trading one against the others, you have a strategic choice, and you need to make it on purpose rather than by accident. The bad version of this conversation skips the question entirely and lets the loudest stakeholder voice in the room – usually whoever has the most direct line to the P&L – settle the discussion. Which is how organizations end up with AI initiatives that hit their cost-reduction targets, miss their customer impact targets, and quietly hollow out the team that was supposed to make the whole thing work.

The good version of the conversation is harder. It requires leaders who can hold three stakeholder views in mind simultaneously, who can resist the temptation to collapse them into one number, and who can defend a decision in plain English to whichever stakeholder ends up on the short end of a particular tradeoff. It also requires honest metrics for each stakeholder – real numbers, not vanity numbers, and not numbers that exist mostly to make the slide deck look balanced. The next article in this series will dig into what those metrics actually look like.

And one more thing, which is worth a brief mention here even though it’s a separate problem. Even a clean three-stakeholder conversation, with honest metrics and a good leadership team, will sometimes produce projects that don’t deliver the value they promised. Not because the value model was wrong, but because the project itself wasn’t aligned with the actual goals of the business in the first place. That’s a different problem, and a harder one, and we’ll come back to it at the end of the series.

What’s coming next

The articles ahead each take one of these conversations and dig in – shareholders, employees, customers, in that order. After those three, a fifth article looks at how the metrics work together when you’re holding all three conversations at once. And the sixth article tackles the harder problem hiding behind the easier one: even with a clean value story, projects fail when they aren’t aligned with the actual goals of the business. We’ll get to that one at the end.

So, the next time someone in a budget meeting asks “okay, but what’s the value?” – you have a better answer than the silence in the room. Three conversations, three stakeholders, none of them optional.

It’s a harder answer than the spreadsheet wanted. But it’s the honest one, and the leaders who can hold all three conversations at once are the ones whose AI investments will actually deliver.

Want to think through your own AI value conversations? Sign up for the JazzAI mailing list – practical frameworks for mid-market leaders making technology decisions without a seasoned CIO in their corner.

Related Articles

• Where’s the Value in AI? (BCG, October 2024) – The earlier BCG report on AI adoption that lays the groundwork for the 2025 leader-laggard gap research, with industry-by-industry maturity rankings.
• The State of AI: Global Survey 2025 (McKinsey, 2025) – Annual benchmark on AI adoption across nearly 2,000 organizations, with the 5.5% high-performer cohort and where their advantage actually comes from.
• Is Generative AI Creating Shareholder Value in Publicly Listed Companies? (MIT Sloan, with Bain) – Direct treatment of the GenAI-to-shareholder-value question, useful companion to Article 2 in this series.

Recommended Books

• The Lean Startup by Eric Ries – The build-measure-learn loop applies directly to AI initiative justification. Validated learning beats the soft-dollar hand-wave every time.
• Competing in the Age of AI by Marco Iansiti and Karim Lakhani – The HBR Press canonical text on AI strategy and value creation. Heavier than this article but useful for leaders who want the full operating-model argument.
• Don’t Think So Much by Jim MacLennan – The practical mindset for tech-aware leadership, including the leadership conversations that resist getting collapsed into spreadsheets.