Article · 24 August 2026
Ninety-nine per cent of Fortune 500 HR chiefs call AI strategically important. Half are not confident their managers can guide it. Britain spent forty years with electricity and no electricity system for exactly the same reason.
There is a number in Gallup's latest workplace research that should give any senior team pause, and it is not the one about AI.
Between February and March this year, Gallup surveyed 102 chief human resources officers from global Fortune 500 companies through its Global CHRO Roundtable. Ninety-nine per cent said AI was somewhat or very important to their organisation's strategy. Fifty per cent said they were not very confident, or not at all confident, in their own managers' ability to guide employees on using AI at work.
That is the whole problem in two sentences. Near-total conviction at the top. A coin toss's worth of confidence in the people who actually have to deliver it. And nobody, so far as the research shows, has asked those people whether they have the room.
Gallup's Q1 2026 workforce study, which surveyed 23,717 employed adults in the United States, found that rolling out AI does not reliably make a workplace better or worse. It makes it move. Where AI has been implemented, 24 per cent of employees say their culture has improved over the past year and 25 per cent say it has worsened. Almost a dead heat. In organisations that have not adopted AI, 59 per cent say nothing much has changed at all.
What separates the two directions is not the tool. It is the manager.
Employees who strongly agree that their manager champions AI are eight times more likely to say AI has transformed how work gets done than those who do not: 33 per cent against 4 per cent. On culture, employees who strongly agree their manager actively supports their team's AI use are more likely to say things have improved over the past year, 31 per cent against 21 per cent, and three times more likely to say the culture has improved a lot, 9 per cent against 3 per cent.
So the strategy is sound. Put a capable, confident manager in front of the technology and the technology lands. The trouble is what that sentence quietly assumes.
Gallup, to its credit, says the quiet part out loud. Its own research shows managers already carrying expanding responsibilities and widening spans of control, and it warns that asking them to lead AI transformation without first considering their capacity risks turning a priority into "another competing demand".
That is a polite way of describing something I see in almost every organisation I work in. A genuinely good idea gets handed to the most overloaded people in the building, and nothing comes off the plate to make room for it.
Britain led the way. In 1881 the small Surrey town of Godalming became the first place in the world with a public electricity supply. We had the inventors, the engineers and the head start, and yet for more than forty years progress at national level barely registered.
The issue was never the technology. Each town, council or private company simply built its own separate system. By the early 1920s there were more than six hundred electricity supply undertakings across Britain. They ran at different voltages and different frequencies, and some areas were still on direct current. A machine wired for one town's supply quite literally could not be plugged in twenty miles down the road. Power stations sat half idle in one borough while a neighbouring one ran short, because there was no way to move electricity between them.
Britain did not lack electricity. Britain lacked an electricity system.
Parliament saw the problem. The Electricity (Supply) Act of 1919 created a body of Electricity Commissioners and asked them to encourage the industry to combine voluntarily. Seven years of encouragement produced three joint electricity authorities and four advisory boards with no effective powers. When the replacement bill was put to the Lords in November 1926, the government's own summary of that period was blunt: the policy of persuasion had been given a good trial, and it had failed. Historians of the industry call the years from 1919 to 1926 the diagnosis of failure.
Britain did not lack electricity. Britain lacked an electricity system.
What broke the deadlock was not a better turbine. In January 1925 the Cabinet asked Lord Weir to review the whole position, and his committee proposed what he called a national gridiron: connect the most efficient power stations with a high voltage network, standardise the frequency, and give somebody the actual authority to make it happen. The Electricity (Supply) Act 1926 followed that December. The Central Electricity Board was created in February 1927. The grid was built between 1929 and 1934 at 132 kilovolts and 50 hertz, running over four thousand miles of line and crossing sixty rivers. It came in on time and on budget at £26.7 million, began operating in 1933, and ran as a single national system from 1938.
The results were not marginal. By 1935 just twenty-eight power stations were supplying the majority of the country. The system load factor, stuck between 28 and 30 per cent through the 1920s, reached 38.8 per cent by 1940. The real price of domestic electricity fell 39 per cent between 1920 and 1940, while gas rose 58 per cent and coal rose 16 per cent. Electricity sales per head of population grew 6.4 times between 1921 and 1938.
Forty years of a transformative technology producing almost nothing at national level, followed by a decade of extraordinary gains, and the difference between the two was not the equipment. It was that somebody was finally given the time, the authority and the mandate to design the system the equipment sat in.
Most organisations right now look a great deal like Britain in 1924.
The technology is genuinely in the building and people are using it. But they are using it the way Godalming used it, locally and in isolation. One team has built something clever. Somebody in finance has a set of prompts nobody else has seen. Two departments are quietly paying for different tools that do the same job. Everyone has wired their own patch at their own voltage, and none of it connects.
This is what the Brookings economist Mark Muro described, commenting on the Gallup findings, as a bring your own AI approach. His point was that most people using AI at work are not doing so as part of any firm wide push. It is freelancing. It probably does improve the standard of work an individual can produce, and that is an entirely different thing from improving the productivity of the organisation.
That is precisely the gap between having electricity in six hundred towns and having a grid.
And the only people positioned to turn local, private use into something that works across an organisation are the managers. They are the ones who can see that two teams are solving the same problem twice, that the report now generated in half the time still gets re-checked by hand by somebody downstream, that the handover between functions is where the days are actually going. They are the interconnection. They are also the people with the least uncommitted time in the entire organisation, which is exactly why the connections never get made.
Gallup's engagement data makes the capacity question concrete. Manager engagement globally fell from 27 per cent in 2024 to 22 per cent in 2025, a five point drop in a single year and nine points down since 2022. Managers have all but lost the engagement premium they once held over the people they lead. The layer being asked to carry the largest organisational change in a generation is the layer that has been quietly emptying out.
The good news is that most organisations are already doing something. In the same CHRO survey, 78 per cent said they encourage peer learning and experimentation with AI, 62 per cent are building centres of excellence or internal AI champions, and 57 per cent are providing AI training specifically for people managers. Only 12 per cent reported no formal action at all.
The gap is not effort. It is what the effort is aimed at.
Start with capacity, not capability. Before a single training session, an organisation has to answer one honest question: what comes off this manager's plate to make room for this? If the answer is nothing, the initiative is not a priority, it is an addition, and it will be treated as one. Protected time is a more powerful intervention than any course.
Make the manager's own work the first pilot. Not the team's workflow, not the customer journey. Their own inbox, their own reporting, their own preparation. Someone who has personally felt the relief of winning back ninety minutes on a Friday afternoon becomes a credible advocate. Someone who has only been briefed on the benefits becomes a broadcaster of somebody else's message, which every team can smell instantly.
Teach judgment rather than tools, because the tools will change. The model in front of your managers today will be superseded within a year, and if you have trained people in a product you will be retraining them by next summer. What does not go out of date is the thinking underneath. What is this actually for. What is worth checking. What must never leave the room. Where does a human being have to stay in the loop, no matter how good the output looks. That is behaviour, not software, and it transfers to whatever arrives next.
Give somebody the frequency. Standardisation was the unglamorous heart of the whole grid project, and it is the part organisations skip. Which tools, which data, which uses are off limits, what good looks like, where the shared prompts and workflows live. None of that emerges from six hundred people experimenting politely in parallel. Somebody has to be given the job, and the authority that goes with it.
Count it, or admit you do not mean it. This is where the data gets uncomfortable. While 78 per cent of those CHROs encourage experimentation, only 26 per cent embed AI skills into job expectations or performance goals, and just 8 per cent tie AI adoption to performance incentives. Encouragement without expectation is a hint. Managers are exceptionally good at reading which organisational messages are real and which are decoration, and they allocate their remaining energy accordingly. Britain already ran the seven year experiment on whether encouragement alone is enough.
Give explicit permission to be bad at it for a while. Adopting anything genuinely new involves a period of visible incompetence, and managers are the group least willing to look incompetent in front of their own teams. Unless senior leaders model the fumbling stage themselves, out loud, the honest experimentation happens in private or it does not happen at all.
Every one of those points is behavioural rather than technical, and that is not a coincidence. Britain did not get a grid because it eventually bought better turbines than everybody else. It got a grid because somebody was finally given room to look at the whole country and ask what this would be if we designed it as one thing rather than six hundred. That is a leadership act, performed under pressure, by people who had the authority to think.
Forty years is what it costs to skip that step. We ought to be able to move faster this time, but only if we stop treating AI as something to be installed and start treating it as something to be led. And the people we are asking to lead it are already running at capacity, in a job that has quietly doubled in size while nobody was measuring.
Give them room first. Everything else follows.
I write about the behaviour that decides whether change actually lands in an organisation. My first book, Rebuilding the Village, is out in early 2027 and looks at what happens to people when the structures around them stop holding.
Whether it's for you, your team or your organisation, the first conversation is confidential, informal and without obligation.
Get in touch