Your CFO walks in. January 2026. The question she carries — What’s the return on the $50 million we’ve spent on AI over the last two years? — lands harder than it should because the answer is suddenly, undeniably required. Not theoretical. Not exploratory. Measurable. Now.
This isn’t hypothetical. It’s happening. A year ago, when we were thinking about AI adoption, time felt elastic. Sixty-five per cent of 1,300 CEOs leading companies over $500 million expected it to take three to five years before seeing actual ROI — a comfortable timeline, a license to experiment, to build without accountability. Today the math has compressed violently. Sixty-seven to sixty-nine per cent of those same executives now believe they’ll see ROI in one to three years instead. Nineteen per cent expect it in six months to one year. That swing from “maybe five years” to “absolutely one year” isn’t a marginal recalibration; it’s the difference between exploration and accountability. Then Morgan Stanley announced it publicly: their AI investment cleared the bar. ROI positive. The proof exists. Someone actually did it.
The efficiency trap
Here’s where the temptation grabs you. You lean hard into efficiency because the math is clean. Marketing teams generate fifty per cent more content. Customer service handles more tickets per hour. Operations tighten. Procurement negotiates harder. Real wins, all of them. But your competitors are doing exactly the same thing — fifty per cent more content, leaner operations, harder negotiations. Efficiency becomes table stakes instantly, and table stakes reward nothing.
A professional services firm discovered this brutally. Their largest client, a major corporation, walked in and said: next year, same work, half the price. Not a negotiation. An instruction. That’s what happens when efficiency is everyone’s baseline. You don’t get points for catching up; you’re just caught up, breathing harder than you were before.
Where the money actually lives
Growth lives there — not doing more with less, but doing things you never had the bandwidth to pursue. Long-tail use cases. Personalisation at scale that would require armies of humans. Markets you couldn’t afford to enter because the unit economics didn’t work. Suddenly they do. An agent changes the equation entirely. You can now pursue verticals, segments, geographies that were purely uneconomical before, because the finance — at agent cost — finally works. That’s where real differentiation hides. That’s where you separate instead of run.
We’ve built portfolios with this in mind: some bets are low-hanging, obvious plays where efficiency gains matter; some are moonshots with asymmetric payoff; most sit in the middle — valuable enough to justify the work, complex enough that most organisations aren’t executing them well. Portfolio discipline, then, becomes the move. Commit enough to efficiency so you’re not last, but concentrate the serious investment where nobody else has figured it out yet. Run the portfolio like a venture capitalist would.
The measurement question
Measurement, though — measurement is the thing that changes everything. Old ROI heuristics don’t fit. You can’t measure AI value the way you measure traditional SaaS adoption. You can’t use the percentage-of-staff-using-the-tool metric because the impact is often structural, not volumetric — it shifts what becomes possible, not just what becomes faster. CIOs we’ve spoken with get this. They’re not asking “Does this fit the old model?” They’re asking “How do we design frameworks that actually capture the value we’re seeing?”
It’s 2026. The CFO is waiting. You need an answer that’s honest, not optimistic: where the AI investment actually created differentiation, where it just moved you in line with everyone else, and where the next dollars go. Growth and efficiency. Portfolio that’s diverse. Measurement that’s real. Those aren’t aspirations anymore.
Signal42 | Beyond the hype. Into the value.