The Future We Choose

Artificial intelligence will help us produce more. The challenge is deciding what we want that productivity to deliver.

We can use it to produce today’s outputs with fewer people, reducing costs and increasing returns.

Or, we can use it to expand what people are capable of: solving harder problems, pursuing greater opportunities and investing in the next generation of talent.

Both can improve a business’s finances, but they don’t create the same future.

We choose to pursue the greater ambition. We want AI to expand what people can accomplish and who gets the opportunity to accomplish it. That doesn’t mean protecting every existing task or pretending every role will remain unchanged. There is work we should automate, and there will be situations in which a smaller team is the right answer. But there is a difference between removing unnecessary work and treating human contribution itself as something we should progressively eliminate.

Taken to its extreme, a model in which every technological advance becomes another opportunity to remove human input leaves us with an uncomfortable destination: an economy that produces extraordinary value while giving people fewer opportunities to create, direct or share in it. We don’t regard that as a successful society, however impressive it could be in its productivity. We want people to remain crucial participants in economic progress, with the opportunity to contribute more and improve their lives through doing so.

Any profession that depends on expertise must consider how the next generation will acquire it, especially when machines can perform the work that once justified employing a beginner. Any business built on relationships must decide when and how to preserve them, especially when technology offers to mediate every interaction. Every early adopter of AI, therefore, must choose what to do with that advantage whilst their competitors are catching up.

Through what we build, who we employ, how we organise work and what we reward, we participate in creating it.

These are the three principles we, as a collective, choose to build around.

The disappearance of a junior task shouldn’t have to mean the disappearance of a junior career.

Yet that is precisely the decision companies may take when they assess the value of a junior colleague solely through the work they can produce today. If an experienced professional with AI can complete the work that once occupied several less experienced colleagues, the immediate commercial calculation may suggest that the next junior hire is unnecessary.

But that calculation excludes the vital purpose of early-career development. We don’t employ beginners only because experienced people need cheaper help. We employ them because our professions need a way to create their next generation of experts. Simpler work allows someone to contribute while they learn, but their immediate output is only part of the return. In fact, it might be the least valuable aspect. The rest of the return, the greatest value, comes from the judgement, understanding and responsibility that person becomes capable of developing, if their growth is properly nurtured.

Consider the industrial revolution in Manchester, where new technology transformed trade. A machine could easily replace basic manual effort, prompting factory owners to cut workforce entry. Alternatively, forward-thinking employers used the shift to elevate workers, involving them earlier in design, planning, and mastery of complex machinery. The tool changes the work; the employer decides whether it also closes off an opportunity.

The advancements of AI are offering knowledge work that same choice. Rather than keeping juniors occupied with routine work while experienced colleagues tackle the difficult questions, we can bring them into those questions sooner. In software, for example, they can watch an engineer investigate an unfamiliar system, interpret an ambiguous requirement or reject an apparently convincing solution. They can hear the alternatives being considered and understand why the final decision differs from the first answer. These are the elements of the craft that finished, polished code can’t explain.

They also learn how their team works: the processes it follows, the standards it expects and the way colleagues communicate and make decisions. Working alongside experienced people helps them understand the culture behind those practices and become part of it.

The business benefits too; that should also shorten the route into higher-paid work which can deliver a stable team that grows with the organisation and ensures knowledge stays within the team. We can develop more experts who deeply understand their field and the people they serve, not just the output of a tool. That value is passed back to the team in better remuneration; everyone wins.

Evidently, our argument isn’t that AI makes junior professionals as senior as the work they can produce with it. Our stance is that we should use AI to help them become capable sooner.

That requires active participation rather than passive observation. Juniors need to attempt solutions, investigate failures, defend their reasoning and receive feedback. They need opportunities to make decisions of increasing consequence as their understanding develops. Watching an expert matters, but so does discovering what happens when they try the work themselves.

The opportunity here is to design a richer learning experience, with less time spent on work that teaches little and more time spent developing the understanding that makes someone independently capable.

That experience depends on access to people.

A junior working alone with an agent, receiving answers but rarely encountering an experienced colleague’s thinking, is not the development model we want. Experts must make their reasoning visible while the work happens, whether through sitting together, pairing remotely or using tools that expose how decisions take shape.

An office doesn’t guarantee that access, and remote working must not become an excuse to withhold it.

None of this happens without investment. An experienced practitioner can’t spend meaningful time developing another person while facing an unchanged expectation of maximum individual output. Some of the productivity AI releases must therefore fund time for teaching, shared problem-solving and thoughtful feedback. Developing another professional should count as valuable work in its own right, rather than an additional obligation squeezed in around delivery.

We choose to keep creating routes into our professions for people at the start of their careers, with experienced people helping them develop. We intend to give experienced people the time to teach and newer colleagues access to demanding work sooner.

We want agents enabling people, and people working with people.

Our work should serve a human purpose, improving services, systems and outcomes for the people who depend on them. However capable AI becomes, we shouldn’t confuse its ability to perform more of the work with a reason to remove people from the relationships and decisions around that work.

An answer can be technically sound, supported by evidence and internally consistent while still missing something important to the people affected by it. We need analysis, but we also need to decide what matters, whose interests deserve consideration and which compromises we are prepared to accept. Those choices should remain connected to people who understand their consequences.

A responsible professional needs enough knowledge to question an answer, enough authority to reject it and enough context to recognise when it addresses the wrong problem. When something goes wrong, somebody must own what happens next. A person placed at the end of an automated process without clear, agreed accountability does not provide meaningful ownership.

Relationships matter for reasons that extend beyond accountability. A colleague facing a difficult moment in their career needs someone who will listen and take their concerns seriously. A person dealing with a failing system deserves someone who will stay involved until the problem is resolved. A team confronting a hard decision needs enough trust to disagree honestly and still work together afterwards. These are relationships we want to preserve, not inefficiencies we are waiting to automate away.

This should be a given, but we want it to be explicitly said. The ability AI has to simulate a reassuring conversation doesn’t replace human empathy. We value the shared experience through which people learn to trust one another, understand differences and recognise concerns that have not yet been expressed clearly.

Agents can handle routine coordination without replacing those relationships. They can prepare information, remove administration and help people express ideas more clearly. But we do not want the centre of working life to become an exchange between agents while people retreat into separate supervisory roles. We want the time technology releases to create more opportunities for useful conversation, collaboration and the exchange of ideas.

That also requires recognising the limits of human attention. When a draft or proposed solution arrives in seconds, another round of reviewing, questioning and deciding can follow immediately. Less typing doesn’t necessarily mean less effort. We should not turn every saved minute into another demand for concentrated individual production, then add mentoring and collaboration on top. Teaching is demanding work too; it needs room in the day rather than whatever energy remains after it.

We expect professionals to engage seriously with AI, and we expect employers to provide time and support for that learning. People should also feel able to challenge a tool or a proposed use of it without having their judgement mistaken for resistance.

That is the standard against which we want to judge the working practices we build around it.

A business that learns to use AI effectively before its competitors can gain a significant advantage.

It may deliver work faster, reduce its costs or take on opportunities that others cannot yet pursue economically. It can pass some of that advantage into lower prices, retain some as additional profit or use it to expand. The people who invested, experimented and accepted the risks of moving early have a legitimate claim to the returns.

The mistake is assuming that the advantage of being early will last. Where competitors can acquire comparable tools and develop comparable methods, we should expect the market to adjust. Existing businesses will adapt, startups will adopt the new economics and those unable to respond will face increasing pressure. The productivity improvement remains, but what once distinguished an organisation becomes part of what the market expects.

Businesses then compete from a more productive baseline, with different expectations around cost, quality and output. Adoption alone becomes less of a reason to choose one organisation over another, and the question returns to what each can achieve with the capability available to it.

Consider two businesses that secure a similar early advantage from AI. One takes as much of the immediate savings as possible, stops hiring juniors and relies on its experienced people to direct the technology. The other also adopts AI aggressively, but invests part of the gain in recruiting juniors and giving experienced colleagues time to develop them. Initially, the first business may report the stronger margin. The second is accepting a cost it could have avoided because it expects to need the capability that investment creates.

As AI-enabled delivery becomes normal and both businesses seek to expand, their positions begin to differ. One has been developing professionals who understand its systems, know its standards and can take on greater responsibility. It has also been learning how to turn promising beginners into capable professionals using AI. The other must recruit the expertise it has stopped producing, potentially at the same time as competitors that made the same decision.

Recruiting experienced people will always be part of building a team, but an industry can’t meet its future need for expertise by assuming somebody else will train everyone. If enough organisations close their entry routes, those relying entirely on the recruitment market may find the people they need harder and more expensive to hire. What initially appeared to be an avoidable cost can become a constraint on growth.

This is exactly why investing in early-career development is a commercial decision, as well as a commitment to people. The return takes time, and no business can guarantee that everyone it develops will stay. Nevertheless, the ability to develop expertise remains valuable even as individuals move between organisations. A company that knows how to help people become capable has something it can continue to build on; one that only knows how to buy existing capability remains dependent on its availability elsewhere.

Shareholders should benefit from the productivity gains AI creates. They provided capital, accepted risk and have a legitimate claim on the return. But shareholder value is not created only through what can be distributed today. A business that reinvests part of an early AI advantage in people, knowledge, new services and future growth may create a more valuable enterprise tomorrow. The question is therefore not whether shareholders should benefit, but how we balance immediate returns with investment that can compound those returns over time.

For us, that means reinvesting part of today’s gains in people, knowledge and future growth, alongside a fair return to owners and shareholders. AI adoption can give us a head start. What we build during that period will determine how useful that head start becomes.