AI Without Strategy: Why the Hardest Part of an AI Initiative Is Knowing What to Accelerate

The cost of building with AI has collapsed. Models are cheap, APIs are everywhere, and almost any team that wants to ship something within a quarter can do it. Boards are asking about AI. Vendors are pitching AI. Analysts are running pilots.

And yet, the more conversations we have with operators across PE, corporate development, and operating companies, the more obvious it becomes that the hard part of AI is not the technology.

The hard part is knowing what to point it at.

The companies pulling away from their peers are not the ones with the most pilots. They are the ones with a clear enough strategy that they can name, with conviction, which parts of the business should be accelerated and which should not. The companies struggling are usually the ones launching AI initiatives in lieu of having that clarity.

AI is an accelerator. If you do not know what you are accelerating, or if your leadership team disagrees about it, AI will quietly carry you in the wrong direction faster than ever before.

The 95% Problem Is a Strategy Problem

The headline number from MIT NANDA’s State of AI in Business 2025 report has become famous for the wrong reason. The often-quoted finding is that 95% of enterprise generative AI pilots deliver no measurable P&L impact, despite an estimated $30 to $40 billion in enterprise investment.¹

Most of the coverage frames this as a technology problem, or a model quality problem, or an integration problem. The MIT researchers themselves are clearer about the cause. The issue is what they call a “learning gap” between tools and organizations: most pilots are not anchored to specific workflows, and most organizations are not clear enough about which workflows actually matter.¹

BCG’s 2025 research arrives at a similar conclusion through a different lens. Across the companies they studied, only 5% are creating substantial value at scale from AI, while roughly 60% generate no material value despite significant investment.² The “future-built” companies that capture nearly all the gains share a specific trait: a multi-year vision driven by the CEO with concrete numerical targets, not a portfolio of disconnected experiments.²

McKinsey’s data tells the same story from the other side. 88% of companies use AI in at least one function, but only 39% see any impact on EBIT, and most of that impact is under 5%.³ The roughly 6% of companies that McKinsey classifies as AI high performers, those attributing 5% or more of EBIT to AI use, share a specific behavior: they are nearly 3 times more likely than their peers to have fundamentally redesigned workflows around AI, rather than layering it onto whatever already exists. Workflow redesign was the single attribute most strongly correlated with EBIT impact across the 25 organizational practices McKinsey tested.³

The pattern is consistent across all three studies. The differentiator is not which model you picked. It is whether the AI initiative is connected to a strategy clear enough to identify what should change.

What Acceleration Looks Like Without a Direction

The most public example of this in the last 18 months is Klarna.

In 2024, the buy-now-pay-later company announced that its AI assistant, built with OpenAI, could do the work of 700 customer service agents. Resolution time dropped 82%. Two-thirds of customer conversations were automated. The cost savings looked, on paper, transformative.⁴

Then the second chapter played out. By mid-2025, customer satisfaction on complex disputes, fraud reports, and emotionally sensitive interactions had degraded. Repeat contact rates climbed. CSAT and NPS scores told a story the volume metrics had hidden. Klarna began publicly walking back its AI-only positioning and announced it was rehiring human agents in an “Uber-style” model focused on flexibility and quality.⁵⁶

The CEO, Sebastian Siemiatkowski, said it directly:

Cost unfortunately seems to have been a too predominant evaluation factor. What you end up having is lower quality.⁶

And later:

We went too far.⁷

It would be easy to read this as a customer service AI failure. It is not. It is a strategy failure that AI accelerated.

If Klarna’s strategic priority was operational cost reduction, the AI deployment was on-strategy and brand damage was the price. If the strategic priority was customer trust during a high-growth period running into an IPO, the AI deployment was off-strategy and the cost reduction was the wrong target. The company appears to have not had clarity on which of these was actually true. The AI executed against the louder of the two voices in the room, and only later did the trade-off become visible.

Klarna is not unique. The pattern is now common enough that examples have become a genre:

Air Canada deployed a customer service chatbot that gave a grieving passenger incorrect information about bereavement fares. The airline was eventually forced by a Canadian tribunal to honor the chatbot’s promise, with the tribunal ruling that the company is responsible for everything its chatbot says, regardless of what its other policy pages state.⁸
New York City’s “MyCity” chatbot, launched to help small business owners, instructed business owners to operate in ways that violated city law, including telling shop owners they could refuse cash and telling landlords they could discriminate against tenants using rental assistance.⁹
IgniteTech, an enterprise software company, mandated “AI Monday,” where staff could only work on AI projects every Monday for a year. The CEO ultimately replaced nearly 80% of the staff because they would not adopt AI fast enough, an acceleration measure that may or may not have been pointed at the right outcome.¹

In each case, the AI tool worked. It did exactly what it was deployed to do. The failure was upstream. The strategic question of what should this team be optimizing for, and is that the same thing the AI is optimizing for was either never answered or never aligned on.

Misalignment at the Top Is the Real Bottleneck

McKinsey’s 2025 AI in the Workplace report surveyed 3,613 employees and 238 C-level executives. Their conclusion was direct:

Employees are ready for AI. The biggest barrier to success is leadership.¹⁰

The headwind is not technology adoption. It is leadership alignment on what the technology is for. McKinsey’s authors put it plainly: securing consensus from senior leaders on a strategy-led AI roadmap is not simple, and the process cannot be oversimplified or assumed.¹⁰

A separate McKinsey analysis of board-level AI governance found that only 39% of Fortune 100 boards have any formal AI oversight, and 66% of directors report limited or no understanding of AI at all.¹¹ At the same time, companies whose boards are AI-fluent outperform peers by 10.9 percentage points in return on equity, while those without fall 3.8% below their industry average.¹¹

McKinsey introduces a useful concept here: every organization has an “AI posture,” whether they have named it or not. A company’s posture is shaped by two questions:

Is AI primarily for optimizing the existing business, or expanding into new products and markets?
Should AI be embedded broadly across the enterprise, or applied to selective high-ROI use cases?

Companies that have not answered these questions tend to drift from pilot to pilot with no strategic center.¹¹ The technology team picks projects based on what looks technically interesting. The functional teams pick projects based on what feels urgent that quarter. The CEO communicates AI is “a priority” without naming what it is a priority for. Three years later, the company has spent millions, generated impressive demos, and moved nothing on the P&L.

This is the gap that no model upgrade can close.

What Strategic Clarity Actually Looks Like

The companies that are getting AI to work are not unusually advanced technologically. Most of them have access to the same vendors, the same models, and the same internal capabilities as their peers. What they have done is the harder work upstream.

A few patterns show up consistently across the high performers:

The strategy is named, written down, and shared. Leadership can articulate the two or three things the company is trying to do over the next 18 to 36 months. AI initiatives get prioritized against those, not against generic categories like “automation” or “efficiency.”
There is alignment on the trade-offs. When AI accelerates one thing at the expense of another, leadership has agreed in advance which side wins. Klarna’s experience shows what happens when this is not made explicit before the deployment runs.
AI initiatives are tied to a specific business outcome, not a technology metric. “Reduce CIM screening time from 4 weeks to 4 days for our PE clients” is a strategy-aligned objective. “Deploy generative AI in deal sourcing” is not.
Leadership commits to redesigning the workflow, not just adding AI to it. McKinsey’s data shows the highest-performing organizations are nearly 3 times more likely to redesign workflows around AI rather than bolt it on top of existing processes, and that redesign is the single attribute most correlated with EBIT impact.³

For executives in M&A specifically, this is exactly where things get expensive when strategy is fuzzy. AI in M&A is uniquely well-positioned to act as an accelerator, because the work is information-heavy and judgment-driven. But it is also uniquely well-positioned to amplify a confused thesis. A team that is unsure whether it is pursuing thematic roll-ups or geographic expansion will end up running AI sourcing initiatives in both directions and getting noisy results in both. The accelerator does its job. The direction was wrong.

This is the problem sc0red Services was built to solve—for PE firms and portfolio companies, not only deal teams.

Where sc0red Fits: Strategy Before Acceleration

sc0red Services is strategy-first consulting for mid-market private equity firms and their portfolio companies. We do not start by selling AI or software. We help leadership name what they are trying to accomplish, where AI can genuinely move the needle, and what should not be accelerated.

The way this works in practice:

Listen. We start by understanding your thesis, your portfolio context, and your value-creation goals—hold period, exit timeline, and operating priorities—not by pitching tools.
Diagnose. We assess strategy alignment, technology maturity, and AI readiness across the firm or portco, and surface misalignment before budget is committed.
Design. We build prioritized roadmaps that connect technology decisions to business outcomes: workflow redesign, custom development, sc0red platform where it fits, or an explicit decision to wait.
Deliver. We implement alongside your teams and measure against agreed outcomes—not demo metrics.

The acceleration only works when strategy has been made explicit first. Firms that arrive with aligned leadership and a clear thesis move faster and waste less. Firms that arrive with conflicting priorities or a vague mandate discover, sometimes uncomfortably, that AI faithfully accelerates in three directions at once.

That is not an AI problem. It is the same strategy problem the McKinsey, BCG, and MIT data describe, made visible faster.

When an engagement calls for software from the sc0red platform, we recommend it. When it calls for custom development, we build it. When the right answer is a process change rather than a technology purchase, we say so. The point is not to deploy more AI—it is to deploy the right acceleration for the strategy you have agreed on.

If this sounds like the conversation your firm or a portfolio company needs to have, book a consultation with sc0red Services.

Everybody can deploy AI now. The question is whether you can name the thing it is supposed to make better.

The Era of “AI as a Side Project” Is Ending

The companies that win the next decade will not be the ones that adopted AI first. By any reasonable measure, that race is already over. They will be the ones whose strategy was clear enough, and whose leadership was aligned enough, that AI could be pointed somewhere specific.

This is the inversion that the data is now making unavoidable. For most of the last three years, the dominant question was how fast can we deploy AI? For the next three years, the question shifts to what, exactly, are we accelerating, and have we agreed on it?

For M&A teams specifically, this means doing the strategic work that gets quietly skipped when the pressure is to “do something with AI.” The thesis. The non-negotiables. The trade-offs the team has agreed to live with. None of this is new. But it is now the prerequisite for AI to do anything useful, and it is the part most teams are still treating as the easy part.

The question is no longer can AI accelerate this?

The question is what should we be accelerating, and do we agree?

That is where the next decade of competitive advantage is going to be built.

References

MIT NANDA. (2025). The GenAI Divide: State of AI in Business 2025. Coverage in Fortune. https://fortune.com/2025/08/18/mit-report-95-percent-generative-ai-pilots-at-companies-failing-cfo/ ↩ ↩² ↩³
BCG. (2025). Are You Generating Value from AI? The Widening Gap. https://www.bcg.com/publications/2025/are-you-generating-value-from-ai-the-widening-gap ↩ ↩²
McKinsey & Company. (2025). The State of AI: Agents, Innovation, and Transformation. https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai ↩ ↩² ↩³
Twig. (2026). Klarna AI Assistant: How It Cut Resolution Time 82%. https://www.twig.so/blog/how-klarna-is-revolutionizing-customer-support-with-ai ↩
Customer Experience Dive. (2025). Klarna Changes Its AI Tune and Again Recruits Humans for Customer Service. https://www.customerexperiencedive.com/news/klarna-reinvests-human-talent-customer-service-AI-chatbot/747586/ ↩
FinTech Weekly. (2025). Klarna Reverses Course on AI Customer Support, Resumes Human Hiring. https://www.fintechweekly.com/magazine/articles/klarna-hires-customer-service-after-ai-pivot ↩ ↩²
MLQ. (2025). Klarna CEO Admits Aggressive AI Job Cuts Went Too Far. https://mlq.ai/news/klarna-ceo-admits-aggressive-ai-job-cuts-went-too-far-starts-hiring-again-after-us-ipo/ ↩
Pinsent Masons. (2024). Air Canada Chatbot Case Highlights AI Liability Risks. https://www.pinsentmasons.com/out-law/news/air-canada-chatbot-case-highlights-ai-liability-risks ↩
CX Today. (2025). 3 Times Customer Chatbots Went Rogue (and the Lessons We Need to Learn). https://www.cxtoday.com/contact-center/3-times-customer-chatbots-went-rogue-and-the-lessons-we-need-to-learn/ ↩
McKinsey & Company. (2025). Superagency in the Workplace: Empowering People to Unlock AI’s Full Potential at Work. https://www.mckinsey.com/capabilities/tech-and-ai/our-insights/superagency-in-the-workplace-empowering-people-to-unlock-ais-full-potential-at-work ↩ ↩²
McKinsey & Company. (2025). Elevating Board Governance Through AI Posture and Archetypes. https://www.mckinsey.com/capabilities/mckinsey-technology/our-insights/the-ai-reckoning-how-boards-can-evolve ↩ ↩² ↩³