Successful AI adoption needs workers in the loop

If AI is not already in your sector, it is on its way. The risks of falling behind are considerable. They range from loss of relevance for an entire business to the bleeding away of thousands of jobs and livelihoods. The full effect of the coming changes is still unclear but the upheaval will be far reaching. 

Not all companies will be able to adapt. Incumbents invested in legacy technology or those that exist to facilitate old-form workflows will find it harder. As happened with the shift to cloud computing, the way we do business is set to change so much that companies tied to old, fixed assets will no longer be viable. 

The earlier generation of digital natives has revolutionised consumer services such as banking and financial management. Functions like marketing have been transformed. 

Ignoring AI is not an option. As the technology becomes more capable, companies will have to work out how they can remain relevant. Their workforce must be fully briefed too.

For a business to stay competitive it will have to do more than simply buy a few AI tools or encourage workers to experiment, although both of these will be necessary. Companies must rethink what AI will do to their relevance. Their staff will have to do the same: some jobs will become obsolete but new roles will be created. 

This report will look at how companies can prepare both themselves and their people for the changes to come.

AI has huge potential: will we react fast enough?

A 2023 UK Department for Education report said 10 to 30 per cent of jobs could be automated. These included routine roles such as clerical work in the finance, insurance and legal sectors. It also said people with a higher level of education would be more exposed. 

Other researchers consider the effect of AI by task rather than by role and say that portions of any job could be automated. 

“Jobs are definitely going to be transformed,” says Stephan Meier, the chair of the management division of Columbia Business School, New York. “That means certain tasks are going to be eliminated or automated. In an ideal world that frees up people to do something different within that job category.” 

Meier, who wrote The Employee Advantage: how putting workers first helps business thrive, points to how ATMs took on the work of bank tellers. Customer-facing employees now focus on advisory roles. “Before ATMs [they] were handing out cash and taking deposits. Now [they] actually have to do light financial advice.”

That evolution, with skill-level and organisational changes, took years to unfold. Developments with AI are likely to be far faster, causing greater disruption as companies and employees try to keep up with new skills requirements. 

Meier says: “If you have 10 years to reskill, or 15 years, that’s quite a different story than if we’re talking about three, or even five years . . . the transition period can be very, very painful.”

The lack of dynamism and agility in bureaucracies — in education in particular — concerns him. These organisations may be unable to change fast enough, and this includes adjusting the syllabus to produce graduates with the right skills. 

Meier says: “There will be new work, new jobs and new tasks, but are people ready to do those?”

Jobs are going to change: skills that are scarce will attract a premium

Some jobs will be affected more than others. While routine tasks will be the first to be automated, skilled or knowledgeable workers could lose out too. A report from the National Bureau of Economic Research in Massachusetts, which looked at customer service, said agentic AI assistance could lift the performance of a novice worker above that of colleagues with greater knowledge and more experience. 

This makes sense: with AI, the ease of access to information means that knowledge and learning are no longer differentiating factors. 

Mainstream translation, for instance, can be carried out as competently by an app as a human, so in a commercial setting a second language will have reduced value. But one aspect that AI cannot yet be trusted with is risk. This means that translators who specialise in interpretation in sensitive situations will still be in demand. 

Meier says: “[Translation] jobs where all the tasks can be automated will go away” but translators for presidents will still be in demand. 

Sangeet Choudary, the author of Reshuffle: who wins when AI restacks the knowledge economy, highlights constraints such as scarcity of skill when trying to determine which roles might disappear. Only skills that remain both scarce and critical will retain their premium. He says the task of typing did not go away with the advent of word processors, but when the cost of editing vanished, so did expert typists. 

Choudary says: “If a skill is scarce, then someone who carries that skill is in a position to manage that constraint. The more difficult it is to acquire a skill, the more that person can charge a premium.” 

Companies and individuals need to look at which skills will become more easily accessible as part of a model. This will influence investment in labour and training.

Debate still swirls around which skills will be useful for workers in a world with AI. Will “people skills” be more in demand? This is possible, especially in customer-facing roles. An example given by Meier is a financial adviser who no longer needs to collate and interpret the news and stock market moves — but no machine can talk to the client and understand their financial concerns.

He says: “No algorithm will have written into its dataset that a client might be worried about a specific situation, such as a sick parent or child, or that they might be considering a big life transition.” 

Everyone can access AI. The winners will be those who use it differently

AI will hugely change how companies and their workflows are organised. Choudary says businesses should ask these fundamental questions: what is their “right to serve the customer”, and how can they do this better than the competition? 

The next question is “does my playing field change?” AI gives no competitive advantage when everyone can access the same tools. Choudary says: “I need to both have a new right to serve the customer and a new way to win against other players.” 

He says that in the early days of social media, entrants relied on people’s connections to keep them bound to a platform and content was delivered to followers based on networks that were generally sticky. TikTok upended that, delivering content based on what each user tended to watch. Suddenly even someone with no followers could gain exposure, an approach that Instagram and others have adopted. 

The key is to understand how AI flipped the concept of who a business serves and how best to do it. There are other examples of how service delivery changes with new technology. Email made sharing documents easier than by post; the cloud took this further by enabling real-time collaboration rather than requiring a fixed digital file to go back and forth between team members. 

Choudary adds: “If you just leverage AI as a tool, you’ll just speed up existing workflows. What you really have to think through are the new capabilities offered by AI: can I reimagine how my industry works [by using] those capabilities?”

Jobs so far unaffected — workers use AI most effectively

Many researchers point to how little AI has affected the jobs market so far. 

An analysis of the US labour market in August by the Economic Innovation Group, an American bipartisan public policy organisation, found that no occupation theoretically highly exposed to AI had experienced noticeably worse employment trends. 

People entering the jobs market tend to be more vulnerable to marginal changes but even here EIG found no evidence that appointments exposed to AI were more precarious. 

Similarly in the sectors that use AI the most — data processing and publishing — the effect was more of a “brake than a meaningful threat”.

An October report from the Yale University Budget Lab and the Brookings Institution concurs, finding that the labour market has not experienced a “discernible disruption” since the release of ChatGPT. 

According to Goldman Sachs, AI is “expected to have only a modest and relatively temporary impact on employment levels”. It says overall adoption is low and that even as this changes, making some jobs obsolete, other roles will emerge to offset the decline. 

The investment bank points out that about 60 per cent of American workers are in roles that did not exist in 1940, which implies that more than 85 per cent of employment growth in the past 85 years has been driven by new technology.

Only younger technology workers have felt the pinch so far, Goldman Sachs says. Unemployment in tech-exposed roles is up 3 per cent since the start of the year, higher than for their peers in other sectors.

News reports indicate that AI is beginning to affect the technology sector more broadly, with digital-first companies leading the charge to use AI over humans. In August TikTok said it would dismiss hundreds of its moderators in Britain as it shifted to AI. The company had already cut the number of moderators in Europe. It said large language models were “reshaping its approach” although one union official expressed scepticism, saying the company was more likely to be offshoring.

Recent data reveals a trend in tech job losses caused by AI, according to Challenger, Gray & Christmas, a US recruitment firm. It says that of 62,000 job cuts in July, 10,000 were explicitly linked to AI. Not everyone believes this. Meier says AI could be a scapegoat because tech companies “[probably] hired too many people during the pandemic”. 

Away from media and tech, a report by the Massachusetts Institute of Technology in July found only limited evidence of AI’s effect on jobs. It said there was no consensus on future hiring trends. 

More broadly the research found poor uptake and uninspiring results from companies that had adopted the technology. While 80 per cent of businesses have dabbled with enterprise AI solutions, only 40 per cent have deployed them. Of those that have integrated AI pilots into their workflows, a mere 5 per cent are extracting value. 

MIT said there was a 95 per cent failure rate for enterprise AI solutions, usually caused by a lack of customisation, failure of the technology to learn and inability to integrate into existing systems. 

While enterprise AI might be failing, the technology is more productive where employees use consumer AI tools themselves in what the report called a “shadow AI economy”. This was so marked that the study’s authors said the future of enterprise AI adoption lay with organisations that learn from this shadow use to inform procurement. 

In this respect, circumstances have hardly changed since an FT Future Forum report on AI in 2020 which found that workers’ buy-in and collaboration would yield better results than a top-down imposition of solutions.

Develop a clear plan to manage AI and share it widely

Managing a business and its people through the rapidly evolving AI landscape is challenging. With so much noise around the topic and evidence that hiring practices are being affected, it is understandable that employees feel nervous. A survey by Pew Research in October 2024 found that more workers were worried than hopeful about the use of AI in the workplace. 

However, as the MIT report implied, people are not opposed to AI tools. A survey of 1,500 workers by Stanford University, California, found that people want repetitive tasks to be automated but they hope to retain agency and oversight. It said workers preferred collaboration to automation. 

Other research confirms that AI usage is already widespread when workers control its use. A poll by Gallup in June 2025 said the proportion of US workers using AI at work doubled in two years to 40 per cent, and the frequency of use also doubled. The poll found that leaders who put forward a clear plan for integrating AI better positioned their workforce to work with the technology.

Companies should be more proactive in creating the right environment for AI. So far a lot of deployment of AI has lacked safeguards but these are important to ensure the optimisation of company strategy.

Only 22 per cent of users said their company had articulated a clear plan for integrating AI. Employees said “unclear use case or value proposition” was the most common adoption challenge at the corporate level.

Not all enterprise uses are failing. Meier says AI has proved its value in internal marketplace tools that match employees with project work outside a regular team. This improves job satisfaction by providing employees with new challenges while managers can find workers with the right skills for a particular project. 

“Mastercard started this during the pandemic and . . . unlocked 1mn project hours . . . Now that really helps in terms of motivation because [as an employee] you are not stuck, you actually learn something new.” The company also uses the system to match mentors and mentees.

Another use is in making sure that staff are given work to match their skills, which Meier calls “just-right tasks”. An employee’s profile can then be updated and tasks assigned to prevent them becoming bored. For sizeable teams, AI is better equipped than a human manager to collate such information. 

Whichever tasks companies believe AI will manage, hollowing out a workforce in anticipation of change is not advised. 

While McKinsey has noted a tail-off in hiring in jobs vulnerable to AI, it says reducing hiring now is not a viable long-term strategy. Instead it recommends hiring into reconfigured work processes to avoid a future talent gap. To stay ahead, companies must ascertain which tasks can be automated and which require human skills. 

Forrester Research takes this further. The global advisory firm says the human user’s experience with the technology will decide its success or failure — and that the culture around AI must be human-centric to succeed. It agrees that employee buy-in is essential, as is enabling appropriate reliance (rather than lazy over-reliance) on AI tools.

A question of trust: avoid mealy-mouthed euphemisms

For an AI strategy to be successful, executives should fully understand the problems it is expected to solve. There is no silver bullet. As the Tech for Growth Forum has frequently noted, simply adopting technology and trying to find where it might work is not a strategy. 

Each business should empower its workers to suggest where the technology is best adopted. Many staff already use AI and it is better to involve them than to have no control over how the technology is used. Once a strategy is determined and a pilot scheme has been successful, workers should be educated about the business case. 

In a previous FT Forums intelligence briefing, Ashley Goodall, the author of The Problem with Change, said that while change was inevitable it is unsettling. He said leaders needed to give employees a sense of stability adding that they should keep their language “straightforward, honest and real”. 

Trust is a big factor. Mealy-mouthed phrasing fools no one and risks driving staff insecurity, which will affect retention. If a company intends to use AI to take out costs, it is better to say so than to dress it up with euphemisms. Good communication helps staff to be the architects of their own futures, with the company or without.

Meier believes that the approach to adopting AI is similar to managing change elsewhere. “People resist change because of skill or lack thereof, or will or lack thereof. They are either not willing to change because they’re afraid, or they think they are fine as they are, or they cannot change.” 

He says companies need to be transparent about the bigger vision, especially why change is happening. Workers need to be told about the benefits for them and how they can be involved. They must be fully trained and prepared for the transition.