Why Over 40% of AI Projects Will Fail by 2027

The data about AI return-on-investment is in and it's bleak. Roughly 75% of executives admit their company's AI strategy is written more for show than for actual internal guidance, and nearly half now call their AI adoption a disappointment. McKinsey's latest survey echoes this, stating that only about 39% of organizations can point to any earnings impact from AI at all. Add in Gartner's prediction that over 40% of agentic AI projects will be canceled by the end of 2027, and a pattern emerges.

The irony is that the AI tools usually work well. The trouble is what's underneath: broken enterprise programs wearing AI like a fresh coat of paint on a cracked foundation. While the paint looks great on first glance, the foundation is still cracked. Thise cracks start to show up when budget season rolls around.

Read more to learn how to build a CX or AI business case that survives the budget room, and the year that follows. We'll get into the three numbers a CFO actually funds, the real cost of doing nothing, and the simple structure that keeps a program alive past month seven.

ai-reality-check-chart

Most AI Projects Die in the Budget Room

Here's what makes it maddening: the technology in the projects that get scrapped and the ones that payoff is basically the same. The difference shows up somewhere far less exciting, in the budget room.

Good CX teams lose funding all the time, usually for the same reason: they make the case in the wrong language. The work is solid and the plan is real, but they walk in with satisfaction scores, journey maps, and a heartfelt story about frustrated customers, while the CFO is doing math on an entirely different page. Those scores are genuinely useful for running a program, and almost weightless in a room where every dollar is judged on return.

A CFO cares about the customer too, they're just also asking a fair question your 40-slide deck doesn't answer: what does this cost us today, what will it earn or save, and how sure are we about the timeline? If the best answer you've got is that customers are frustrated by the delivery experience, the money goes to the team that showed up with numbers.

Three Numbers That Beat NPS in the Budget Room

The good news is you're probably already sitting on the raw material for a real financial case. You just have to translate what you know into three numbers a CFO already trusts.

Cost to serve

What is the friction actually costing you right now? Repeat contacts, escalations, manual workarounds, rework, it all carries a price tag. When a customer has to call three times to fix one problem, that's two calls you paid for and didn't need. Add it up across a year and the number gets loud.

Revenue retained

Churn is the one word every CFO speaks fluently. Tie the experience gap to the revenue walking out the door because of it, or the revenue you keep by closing it, and the conversation stops being soft and starts being financial in a single sentence.

Time to payback

A CFO funds certainty as much as size. A smaller return with a credible 14-month payback and a name attached will beat a bigger number with no timeline and nobody accountable. Show the payback period and the return, and you've answered the question before they ask it.

NPS still earns its place for running the program day to day. These three numbers translate that work into the currency the budget actually gets decided in.

The Number Every Failed Case Leaves Out: The Cost of Doing Nothing

Every strong business case includes one number almost everyone forgets: the cost of doing nothing.

Inaction feels free because it never lands on an invoice. It isn't. Every quarter the fragmented journey stays fragmented, the repeat contacts keep coming, the at-risk revenue keeps leaking, and the competitor who fixed their experience keeps pulling away.

Putting a number on standing still does two useful things. It turns the decision from "should we spend on CX?" into "which cost do we want to keep paying?" And it hands the CFO a baseline to measure the initiative against later, which is exactly the kind of accountability that makes them comfortable saying yes in the first place.

Why the Survivors Make It Past Month Seven

Winning the budget is the easy part. The hard part is building a case that's still standing in month seven, when something else catches fire and the discipline quietly slips. A business case is really just one piece of a well-run customer experience program, and the ones that survive tend to share the same simple bones:

Baseline. The current-state number, measured before you start, so the impact is provable later.
Initiative. The specific change, scoped tightly enough that you can actually isolate what it did.
Owner. One person, accountable for the outcome by name.
Benefit realization plan. How and when the return shows up, and how you'll track it against that baseline.
Governance routine. The recurring rhythm that catches drift in days instead of quarters.

That last one is the whole game. It's the line between funded-and-forgotten and funded-and-delivered. If you're building this out, we've put the practical version of each piece in writing: how to set up program governance that actually holds, which KPIs are worth counting, and how to build project scorecards that keep everyone reading from the same sheet.

What It Looks Like When It Actually Works

Enough theory. Here's the pattern in the wild.

A Health Plan: Four VoC Programs Into One

A major health plan came to us running four separate voice-of-customer programs. Four teams, each funded, each staffed, each telling a different story about the same members. We helped consolidate them into one. The result was $2 million in savings and 18 coordinated initiatives in the first year. The savings got the headlines, but the coordination is what actually moved things, because no single team watching a quarter of the journey could ever recommend a fix that spanned the whole thing.

A Wireless Carrier That Got Ahead of the Crisis

A wireless carrier brought us in before they had a crisis, which is the smarter and rarer version of this story. Over two years, the program returned a 93% internal rate of return and paid for itself in 14 months across 12 initiatives. What those numbers don't show is the governance underneath them. The work survived two leadership changes and four shifts in priority on the client side, which is the real test of whether a business case was built to last or just built to get approved.

In both cases the technology mattered less than the discipline around it. A baseline. A number per initiative. A named owner. A routine that kept everyone honest.

The Real Difference Between the 40% and the 60%

So, here's the point to carry into planning season. The AI and CX investments that get scrapped and the ones that pay off will run on roughly the same tools. What separates them is the unglamorous work: translating the effort into the language finance already funds, attaching a real number to every initiative, and building the governance to prove it out. That's the foundation under the paint.

Do that work, and you land in the 60% that survives. Skip it, and you join the 40% waiting to be canceled.

If you're putting together a CX or AI case for budget season and you want it to survive the room, that's the work we do every day. Read the full story behind the health plan consolidation or talk to our team

Joe Piette

EVP, Customer Experience

Speak to an Expert

Talk with our EVP of Customer Experience about how we can help you design and implement customer, employee, digital, or contact center experience strategies that drive results.

Book a 30-minute meeting→

Frequently Asked Questions

Why do so many AI projects fail?

Gartner predicts over 40% of agentic AI projects will be canceled by the end of 2027, driven by escalating costs, unclear business value, and weak risk controls. In most cases the technology works fine. What's missing is a measurable business case, a clear owner for each outcome, and the governance to carry the program past the pilot.

What percentage of AI projects fail or get canceled?

Gartner projects that more than 40% of agentic AI projects will be scrapped by the end of 2027. The measurement gap is just as stark: McKinsey's latest survey found only about 39% of organizations can attribute any earnings impact to AI at all.

How do you build a business case for customer experience?

Start with five pieces: a measured baseline, a tightly scoped initiative, a single accountable owner, a benefit realization plan, and a governance routine to track it. Then translate the work into financial terms a CFO already trusts, like cost to serve, revenue retained, and time to payback.

What CX metrics matter more than NPS when proving ROI?

NPS is useful for running a program, but three numbers carry more weight in a budget room: cost to serve (what the friction costs today), revenue retained (the churn you avoid), and time to payback (how fast the return arrives). These translate customer experience into the currency budgets are actually decided in.

What is cost to serve, and why does it matter for CX?

Cost to serve is the full cost of delivering an experience to a customer, including repeat contacts, escalations, manual workarounds, and the rework that friction creates. It matters because it turns a customer experience problem into a finance number: when a customer has to call three times to resolve one issue, that's two contacts you paid for and didn't need. Lowering cost to serve is one of the most direct ways to show CX return in a budget conversation.

How do you justify CX or AI investment to a CFO?

Make the case in financial language rather than satisfaction scores. Show the cost to serve, the revenue at risk from churn, and a credible payback period with an owner attached. Then quantify the cost of doing nothing, so the decision becomes which cost to carry rather than whether to spend at all.

What is the cost of doing nothing in CX?

It's the recurring, often invisible cost of leaving friction in place: repeat contacts you pay for, escalations, and revenue that leaks out through churn every quarter the problem goes unsolved. Putting a number on inaction reframes the funding decision and gives you a baseline to measure results against later.

What separates the AI and CX projects that succeed from the ones that fail?

Rarely the technology. The programs that survive share a measured baseline, a number per initiative, a named owner, and a governance routine that catches drift in days instead of quarters. That discipline is what carries a program past month seven, when priorities shift and most efforts quietly stall.