Round Two: Your Second Swing at AI Has to Land

The first wave is over. Agents rolled out, some into live workflows. Copilots on every desk. Teams running their own experiments, picking their own tools, working in silos.

Now the bill is arriving, and it doesn't add up. Token costs compound month on month while the productivity line stays flat. The centralised gains the platform pitch promised never showed, because every team built alone and nobody set a baseline. People are confused about the tooling, the process changes, and where they fit in the new shape of their own job. Finance is asking what the spend bought, and nobody has a clean answer.

So the resets have started. Not in theory. In production.

Strike one, by the numbers

MIT's NANDA research found 95% of enterprise generative AI pilots delivered zero measurable ROI, against $30 to $40 billion in spend. S&P Global reports 42% of firms abandoned their primary AI initiative between 2025 and 2026 because they couldn't prove a path to return. Sinch surveyed 2,500 senior decision makers this year and found 74% of enterprises with a live AI customer agent have rolled it back or shut it down. Gartner expects over 40% of agentic AI projects to be cancelled by 2027.

Here's the part that matters. Despite all of this, 98% of enterprises say they will increase AI investment in 2026. The money is coming back. The patience isn't. Boards watched one cycle burn time and budget with little to show. Round two runs on a shorter leash, and everyone at the table knows it.

Why the first swing missed

The failure pattern is consistent, and it has nothing to do with the models.

Most organisations looked at exactly two things: their humans and their AI tools. Connect the two, run the training webinar, done.

But an organisation is at least eight things: people, roles, tasks, skills, processes, tools, vendors, and agents. Every one connected to the others. Tasks are completed by roles, agents, or vendors. Roles require skills and use tools. Processes orchestrate roles. Change one and the effects ripple through all of them.

Looking at two of the eight guarantees the failure we just watched. AI deployed with no organisational context guesses. Humans handed new tools with no guidance, training, or guardrails improvise. And none of it can be measured, because nobody mapped the work before changing it. You can't track progress against a baseline that was never captured.

This is the Human Layer Graph: the full map of how work actually happens. Humans are one node type among eight. Essential, but not sufficient. The first wave deployed into the two visible layers and ignored the six that hold everything together.

Nobody owned it

There was a second flaw, and it sat at the top of the org chart. Ownership was cloudy from the start. A CHRO programme because it changes roles and people? A COO programme because it changes processes? In most organisations the honest answer was all of them and none of them, so decisions stalled, spend fragmented, and no one was accountable for whether any of it landed.

The burn has changed the room. The CFO was always involved. After a year of token bills with no productivity line to match, the CFO is watching everything like a hawk. BCG finds high performers are 1.4 times more likely to have finance enforcing accountability in their AI programmes. That's the new normal.

The market's other answer is titles. IBM's 2026 CEO study found 76% of organisations now have a Chief AI Officer, up from 26% a year earlier, and Chief Transformation Officer is being bolted onto roles at pace. It helps: companies with a CAIO report a 5% higher return on AI investment. But a title is not a map. A CAIO inheriting the same two-of-eight view inherits the same failure, just with clearer blame. Korn Ferry's research shows how deep the gap runs: 42% of CHROs say they're prioritising AI investment, while only 5% of their teams feel prepared to put it into practice.

What's actually landing is different. Success shows up where every team is inside the programme: HR on roles and skills, operations on processes, IT on tools and agents, finance on measurement. And where the work runs as small, targeted programmes with results measured before anything scales. Big swings do land results, but they're far harder to measure, and on a shorter leash what you can't measure you can't defend.

The adoption divide: culture

Then there's the failure mode nobody puts in the steering committee deck: culture.

The first wave's enablement plan was to put everyone onto a copilot at the same time and call it done. No role-specific training, no answer to the only question that matters to the person in the seat: what does this change about my job?

The result is a workforce splitting in two. Some employees jumped at the chance and are compounding gains weekly. Others are slower to adopt, and here's the dangerous part: they're often your most experienced and knowledgeable people. The operators who hold the exceptions, the customer context, the institutional memory that automation runs on. The data shows how wide the gap already is, from both directions. IBM found 86% of CEOs believe their employees have the skills to work with AI, while only 25% of the workforce actually uses it regularly. Korn Ferry surveyed 15,000 employees and found 78% of leaders believe they have AI figured out. Just 39% of their workers agree.

You don't want a workforce of fast AI adopters on one side and your institutional knowledge experts on the other. Fast adopters without deep context produce output that looks like progress. Experts without AI fluency get quietly left behind, and their knowledge leaves with them. Neither half wins alone. Closing the divide starts with seeing it, role by role: who's adopting, who isn't, and which slow adopters hold knowledge the transformation can't afford to lose. Then targeted enablement, not a webinar, and a culture of change driven from the centre.

The winners went to the base layer

BCG's global research puts hard numbers on who's pulling ahead. Just 5% of companies qualify as future-built for AI. They generate double the revenue growth and 40% more cost savings than the laggards, who make up 60% of the market and report minimal gains from the same technology.

What separates them isn't better models. BCG's own conclusion: most value comes from changing how work gets done, not from the AI itself. The leaders redesigned workflows, prioritised a few high-impact moves, tracked financial impact rigorously, and scaled from proven wins. They knew the work before they changed the work.

What round two looks like

Three phases. In order. No shortcuts.

Discovery, properly. A true discovery and ingestion phase at the core. Map the full organisation across all eight areas. Roles are bundles of tasks, so map the work, not the job titles. This is the baseline every later claim of progress gets measured against.

Transform from the base, in parallel. Not one giant pot of spend. A portfolio of small, targeted programmes: agent rollouts, tool consolidation, role redesign, L&D, vendor rationalisation. Each scoped against the map, each with measurable outcomes defined before it starts, every team inside it. Measure, then scale. Gains in one programme compound into the next. That's the centralisation the first wave promised and never delivered.

Traction, continuously. Adoption tracked role by role, so the adopter divide is visible before it hardens. Hours freed, and where they went. Tool and vendor spend against plan. Skill gaps closing or not. ROI as a number that traces back to task-level evidence, the kind a hawk-eyed CFO signs off on.

Strike one is gone. The leash is shorter. The organisations that get a third swing will be the ones whose second swing was measured, and you can't measure what you never mapped.

Round two starts at the base layer.

AI doesn't kill jobs. It frees time. This time, prove where.

---

Sources: MIT NANDA enterprise AI research; Sinch AI Production Paradox report, May 2026; S&P Global; Gartner; IBM Institute for Business Value CEO Study, May 2026; BCG, The Widening AI Value Gap and 2026 AI value research; Korn Ferry Workforce 2025 survey (15,000 global employees) and CHRO research.