What can we learn from the office AI superusers?

Calls to slow the pace of AI development for fear of its potential to destroy humanity are unlikely to resonate with many office workers, most of whom are still primarily using the technology for basic tasks.

Only about a quarter of UK adults use AI tools every day for their jobs, according to a Deloitte survey of 25,000 employees published this month. It found the most popular uses were searching for information, writing emails, fact-checking and creating summaries.

But a smaller group of AI “superusers” have restructured their working lives around intensive use of new tools. Microsoft’s 2026 Work Trend Index report says 16 per cent of AI users fall into this category and identifies three things that define them: they are advanced users of AI agents; they have redesigned workflows to maximise efficiency; and they share their knowledge across their wider organisation.

But what does it mean to class yourself as an AI superuser in the office? Senior corporate executives share their experience.

Feed agents with institutional knowledge

For some bosses, an AI agent acts like a chief of staff. Pamela Maynard, chief AI transformation officer for Microsoft customer and partner solutions, describes hers as holding “institutional context — my priorities, my calendar, what I care about, even my boundaries”. She says the agent uses this “to set me up: shaping my week, surfacing what needs my attention and making sure I walk in prepared”. It also pushes back: “It has actually asked me why I was working on a Sunday afternoon when I should have been with friends or family.”

Angela Tangas, head of the UK marketing agency Oliver, part of AI and digital advertising group Brandtech, has an “Ask Ange” agent, not yet an avatar, but a useful first line of support for her direct reports when she is not available. “It started with some of the most common questions I am asked when speaking to customers, and how I would typically respond to those questions, and where I go for sources of information.” It is a work in progress: ask it something too complex and it “cuts off” and refers questioners to the real CEO.

Redefine what you can achieve

At law firm Lewis Silkin, senior associate Jordan Quartey uses specialist legal AI tool Harvey to gather client-specific files containing the firm’s previous advice and decisions, as well as current email chains. This means he no longer has to ask senior colleagues for advice at an early stage, getting further along in the process on his own by learning what has been done before, and using that context to develop his judgment. “It’s more efficient, and it’s a better way of learning for myself,” he says.

Ivan Zhao, co-founder and CEO at Notion, a collaboration and productivity platform, likens the shift in working patterns to the difference between a marching band and a jazz ensemble. “The marching band is like the old way of working — everyone in lockstep formation, following the same old sheet music, spending a lot of energy just to stay in lockstep.”

A jazz band, meanwhile, “has talented people improvising around a shared sense of where the music is headed”. In his workplace this means agents handle the “boring and repetitive stuff” so staff can “spend their time on the improvisation and fine-tuning”. There are “more long conversations with colleagues, riffing on what we are building, why we’re building it that way, and what we should do next”.

Prioritise human colleagues

Maynard at Microsoft says “AI should give us more room for the human, not less”. Time saved through efficiency gains should be used for “relationships, the mentoring, the messy human conversations no agent can have for me,” she says. “The real risk . . . is that we let the recovered time quietly refill with more tasks.”

While some leaders are keen to cut meetings, Maynard makes them “count for more”. An agent does the pre-briefing so teams can use the time for “actual conversation”. “The recap and follow-ups write themselves, so the time in the room goes to people, not process.”

Zhao says Notion staff have eliminated “boring meetings”, because “an agent can summarise what products shipped this week or what sales were made”. That means “there’s more time for spontaneous meetings, where you realise something over lunch or drop by someone’s desk and chat about what’s on their screen — this is often where the best ideas come from”.

Stop thinking of AI as a time saver

Tangas says she does not measure the benefits from her team of AI agents in “time I’m saving” but how it makes her “more effective as a leader”.

She starts the day with an AI-generated news round-up, prompted and refined to give her news about relevant sectors. “For example, if [a client] has a profit warning, would I have seen that if I didn’t have my prompt? Probably not. It’s enabling me to be much more proactive.”

She then works with agents within the company’s own AI platform to stimulate ideas and prepare for client meetings. “Let’s say before I am going to see an executive — I would say, ‘My profile of the customer looks like this’ and ‘this is what I hope to achieve from the conversation. Can we role-play how that conversation might play out?’”

Leanne Taylor, chief executive of Syspro, a software maker, says how many hours she saves “is the wrong question”. Her company’s AI tool is trained on corporate routines and discussions but also learns how she personally thinks and makes decisions, and what her priorities are. It understands the “language I use when I’m thinking something through, versus when I am landing [at] a conclusion”, she says. This means the technology can help her make better decisions that “reflect how I lead and not a generic version of leadership”.

Take care with ‘bot sitting’

Berry Diepeveen, a senior partner at BCG and leader of the consultancy’s tech and digital practice for the UK, Netherlands and Belgium, says he has “virtual teams working for me day in and day out”.

But he stresses that proper oversight is vital: “If I create a lot of AI-enabled content, but I pass it over the fence, and a human needs to suddenly deal with five, ten, fifteen times the volume of work, then, of course, that bottleneck has shifted.”

Have fun

“The most surprising thing about AI? How much fun it is,” says Taylor at Syspro, “I expected useful and I expected efficient. I wasn’t expecting to really enjoy the way I work more than I did before. But there’s something that happens when the friction drops — you start playing.”