The shift you can feel but cannot measure

If you've been leading a regulated SaaS business through the last eighteen months, you'll recognise this.

The early energy was good. Teams were trying things. Pilots were running. "Have you seen what it can do with this?" moments in every Slack channel.

Now the conversation has changed. The board wants outcomes, not experiments. Customers are asking what your AI is actually doing in their workflows. The CFO wants ROI on the spend, not licence counts. And privately, you're wondering whether your team has actually moved past the playing-around phase.

Microsoft's 2026 Work Trend Index gives the shift a measurable shape. Only 19% of workers sit in what Microsoft calls the Frontier zone — the place where individual capability and organisational readiness reinforce each other. Half the workforce is in the messy middle. One in ten are skilled people working inside companies that haven't kept up. Across the whole sample, 65% fear falling behind if they don't adapt with AI quickly, 45% say it feels safer to focus on current goals than to redesign work, and only 13% say they're rewarded for reinvention when it doesn't produce immediate results.

The cost of getting this stage wrong is becoming visible in trust, retention, and risk all at once.

What I see when teams are caught between phases

I work with the people in the middle of this shift. The teams who've experimented with AI but haven't moved to relying on it.

Three things are almost always missing.

Readiness. Rarely about training hours. Almost always about whether the team can honestly say where they're confident and where they're guessing.

Confidence. Two sides. The courage to use the tools, and the courage to push back when they're wrong. Most organisations underinvest in the second one.

Responsibility. Clarity about who owns the output once AI has helped shape it. In regulated SaaS, that conversation has to happen before reliance, not after.

Microsoft's data sharpens why these gaps persist. When Microsoft tested 29 factors against actual AI impact, organisational factors — culture, manager practice, talent systems — accounted for 67% of the variation. Individual mindset accounted for 32%. The single strongest driver was the organisation's AI culture, around 2.5 times stronger than any individual factor.

Fifteen years working through technology shifts, the pattern is the same. The shift from experimenting to relying isn't a bigger leap than the original adoption. It is a different leap. The first one needed permission. The second one needs preparation.

What it actually takes

Most leaders rush this stage or avoid it entirely. Rushing breaks trust. Avoiding it leaves the organisation in permanent pilot mode while competitors compound real capability.

What's needed is a deliberate move from individual experiments to team agreements. An individual experiment is "I tried this." A team agreement is "we have agreed AI does this part of this workflow, with these guardrails, and this is how we check it works." That difference is the difference between curiosity and capability.

It also takes investment in confidence on both sides. Leaders modelling AI use openly, sharing what worked and what didn't. Teams trained to push back on AI outputs, not just consume them. Quality checks built into the workflow, not bolted on at the end.

Microsoft's findings on manager behaviour are unusually direct here. When managers visibly model AI use, the team sees a 17-point lift in reported AI value, a 22-point lift in critical thinking about AI, and a 30-point lift in trust in agentic AI, and is 1.4 times more likely to use agentic AI at high frequency. Manager modelling is the cheapest, fastest intervention available. Most organisations don't price it that way.

In regulated SaaS, it also takes an explainability layer that lets teams answer "how was this decision made" in front of a board, a client, or a regulator. That layer is becoming the price of reliance.

There is a leader/employee perception gap worth taking seriously. In Microsoft's data, 81% of leaders say their people feel safe suggesting new ways of working with AI; only 67% of employees agree. 78% of leaders say their managers create space for experimentation; 59% of employees do. 21% of leaders say reinvention is rewarded regardless of immediate outcome; only 10% of employees report the same. Leaders systematically think the system is more permissive than it is. The decision to rely on AI is being made on rosier data than the ground truth.

Done well, AI moves from a side experiment to core operating capability. Adoption stops being an effort and becomes a default.

See where your team actually is on the shift

The hard part is that "are we ready to rely on AI" is almost impossible to answer from inside the leadership team. The answer is in the squads, and most leaders only see filtered versions of it.

The AI Risk & Value Scorecard gives you the unfiltered version, specifically through its Adoption and Integration core area. Eight sub-areas, including workflow integration, user adoption, decision integration, behaviour change, operational ownership, user feedback, system alignment, and friction removal.

Each is assessed against evidence and given a 0–4 maturity score. You see whether AI is part of the work or beside it. Whether users are relying on it or reverting. Where ownership is clear and where it isn't. Where friction is dropping people out of the workflow before they finish.

The scorecard pairs that with your Culture and Capability score and your Measurement and Value Realisation score. You end up with a real picture of where you are on the shift from experimenting to relying, and a 90-day roadmap for closing the gaps that matter most.

Find out more at scailwithai.com/scorecard