AI Needs Translators With Operating Authority

Boards want vision. CEOs want realism. Organisations need someone who can turn AI capability into work the business can actually own.

I used to write a lot about AI translators and connectors. More than a year ago, that language was already beginning to appear in the market. MIT and McKinsey had talked about the need for people who could bridge AI and business, and I agreed with the basic idea. But even then, something about the framing felt too small, because translation was too easily treated as explanation: a useful comms layer, someone to make technical work legible, tell a better story, smooth the handover between engineers and executives, and stop everyone looking baffled in the steering committee.

That was never enough. AI does not only need people who can explain technology in compelling language. It needs people who can turn capability into operating decisions, moving between strategy, workflow, cost, trust, governance, proof, and human reality without losing the thread.

A year ago, I put it this way: AI needs translators with a P&L. I still think that was right. But I would sharpen it now: AI needs translators with operating authority.

Not because every company needs another grand AI title. I am not especially interested in titles for their own sake, and the market already has enough impressive-sounding roles floating above unresolved work. The point is to name a missing function. Someone has to turn AI from possibility into work the organisation can fund, govern, scale, or stop.

The BCG CEOs and Boards survey captures why that job is becoming necessary. At first glance, CEOs and boards appear aligned. Both recognise AI as strategically important. Both know it cannot be parked in a technical corner and ignored. But once you look beneath the surface agreement, the split becomes much more revealing.

CEOs worry that boardroom judgement is being distorted by hype, while boards want CEOs to articulate a clearer and more ambitious AI strategy. CEOs are asking for realism. Boards are asking for vision. Both are reasonable, but together they pull in opposite directions.

That pressure becomes acute when pace enters the picture. BCG says roughly 60% of CEOs believe their boards are rushing AI transformation, and that lower board confidence in AI knowledge may be fuelling FOMO. The report also says frustration with the pace of adoption is highest among board members with a more permissive view on governance.

Put simply, the people most anxious about moving slowly may also be the least inclined to demand strong oversight.

That is a dangerous combination if nobody is translating between ambition and operating reality. Board impatience can become pressure for movement before the organisation knows what it can absorb. CEO realism can sound like defensiveness if it is not turned into a credible path forward. Governance can be experienced as a brake if it arrives as late objection rather than part of the design. And AI strategy can become a language that reassures everyone while the real decisions remain vague.

This is where the missing leadership capacity matters. The useful translator is not the person who says, “Let me explain what the model does.” The useful translator says, “If this system enters this workflow, here is what changes: decisions, costs, risks, support needs, controls, and the proof required before authority widens.”

That is a different job. It is not soft translation. It is translation with consequences attached.

In my AI work, this is where the useful conversation usually starts. Not with another abstract debate about whether AI is transformative, and not with a tour of tools. The real question is what the organisation is now prepared to let AI change. The workflow. The decision rights. The cost model. The control points. The human responsibilities. The proof standard.

That is where AI stops being an executive talking point and becomes a business commitment.

The BCG report also says a majority of boards and CEOs believe AI should sit with the executive leadership team, but in practice CEOs still take much of the burden. According to the survey, 47% of CEOs report heading up AI implementation, while fewer than 10% of both CEOs and boards think AI strategy should be led by a dedicated chief AI officer.

That is interesting because it says AI is recognised as a whole-business issue, but still collapses back onto the CEO when accountability gets real.

I understand the reluctance around a dedicated AI leader. A badly defined CAIO role could become a symbolic executive, a vendor magnet, a policy wrapper, or a centralised bottleneck that everyone can blame.

But the opposite mistake is just as serious.

If no one holds dedicated AI leadership capacity, the organisation ends up with distributed involvement but fragmented accountability. The CEO carries the story. Technology carries the tooling. Product carries use cases. HR carries training and anxiety. Risk and Security carries late objections. Finance carries ROI disappointment. Operations carries the exceptions.

Everyone has a piece of the problem, but nobody owns the work of turning AI into something the business can stand behind.

That is where value forms or leaks away. You can have executive enthusiasm without operating clarity. Technical capability without adoption. Adoption without enterprise value. Governance without speed. Speed without legitimacy.

The point is not that every firm needs a CAIO by that exact title. The point is that every serious organisation needs the capacity the title is trying, often clumsily, to describe.

Someone has to know enough about the technology to avoid being dazzled by demos, enough about the business to understand where value actually moves, enough about operations to see where work will break, enough about risk to design boundaries early, enough about finance to make the economics legible, and enough about people to understand when adoption has been poisoned by mistrust.

This is also why AI leadership cannot be left as generic enthusiasm.

Different executive homes will naturally bias AI in different directions. If AI sits mainly with technology, the organisation may over-index on architecture, platforms, tooling, and security. If it sits mainly with finance, the story may collapse too quickly into cost reduction and headcount logic. If it sits mainly with HR, it may become training, change management, and workforce sentiment. If it sits mainly with product, the bias may be toward features, customer experience, and experimentation. If it sits mainly with risk or legal, the organisation may become excellent at naming exposure and poor at capturing upside.

None of those biases are unreasonable. They are the natural gravitational fields of executive roles. The problem is that AI cuts across all of them. It changes cost structures, workflows, decision rights, customer promises, employee trust, data exposure, operating resilience, and proof standards all at the same time.

Treating AI as the property of one function makes the organisation see clearly in one direction and partially everywhere else.

You can see the same confusion in AI leadership job specs. Many read like laundry lists for mythical operators: part technologist, part transformation lead, part product strategist, part risk officer, part change manager, part evangelist, part procurement expert, part workforce therapist.

The breadth is understandable. The problem is that the centre is often missing. The organisation has not decided what this person must make true.

In my work, I increasingly help organisations shape AI leadership roles, CAIO-style mandates, and recruiting briefs, and the first task is usually cutting the unicorn language so the actual translator role can appear: not someone who does everything, but someone with enough operating authority to translate AI ambition into workflow decisions, proof standards, authority boundaries, economic discipline, and work people can trust.

That is why the translator role needs authority, not just fluency. Without authority, translation becomes commentary. The person can explain the gap, name the tension, and probably be proved right later, which is satisfying only if you enjoy watching avoidable confusion mature into expensive lessons.

With authority, translation becomes a way of forcing decisions early enough to matter.

Those decisions are practical, not philosophical. The organisation has to decide what AI is allowed to change, which workflows are being redesigned rather than decorated, where decision rights move, and what remains human-held. It has to know what proof would justify scale, what cost would break the business case, what would damage trust, and who can stop the system before a small issue becomes a board problem.

This is where the trust and economics stories belong, not as side issues, but as direct consequences of the operating choices. If employees believe AI is being introduced to expand what they can do, adoption has one texture. If they believe they are being asked to document their work so it can be extracted, automated, and used against them, adoption has another. If finance prices only licences and ignores compute, retries, supervision, exception handling, integration, audit, quality control, and failure recovery, the business case may look persuasive while the workflow stays expensive.

Accenture’s recent UK report is useful here too because it has been widely misread. I’ve seen a lot of commentary treat it as evidence that LLMs are failing in the enterprise. I think that reading is too thin. The more interesting point is that capability is arriving faster than organisational redesign. The technology is not irrelevant, and model limits still matter, but the deeper gap is between what AI can now do and what organisations are structured to absorb.

The stronger reading of the current enterprise AI gap is not simply that LLMs are weak, although they sometimes are. It is not simply that vendors overpromise or consultants turn uncertainty into oversized programmes, although both happen. It is that organisations are struggling to convert capability into impact because the work has no clear home.

Accenture’s UK report supports the same reading from another angle. It argues that AI capability is advancing faster than economic impact, and shows employees adopting AI more quickly than organisations are redesigning the workflows and systems around them. It also frames the main barriers as organisational and human rather than primarily technological: how work is structured, how decisions are made, how people are equipped, and how trust is built.

That should be encouraging rather than depressing. If the only problem were that the technology was useless, there would be little to do. But if the problem is conversion, then serious leadership can change the outcome.

Organisations can redesign workflows. They can create proof standards. They can price supervision and exception handling properly. They can show employees how AI will expand their range, reduce low-value friction, and make previously impractical work possible, instead of leaving them to assume the whole story is extraction and replacement. They can build controls that make speed legitimate rather than reckless. They can stop treating governance as the department of “no” and start treating it as part of how delegated work scales.

That is the version of optimism worth taking seriously. Not “AI will transform everything because the demos are amazing.” Something plainer and more useful: AI can transform a great deal, but only when leaders make the conditions for transformation explicit.

The translator with operating authority sits exactly there, between promise and condition. They are close enough to the board to understand ambition, close enough to the CEO to understand accountability, close enough to the work to know where reality will bite, and close enough to proof to stop the organisation lying to itself.

They do not need to own every system or approve every decision. That would be absurd. But they do need enough mandate to force clarity before momentum hardens into operating debt.

This is not a soft role. It is not internal comms with a better vocabulary. It is not “the AI person” wandering around with a slide deck and a gift for metaphors, though I remain strongly in favour of a good metaphor when the room deserves one.

It is a leadership role for an awkward phase of enterprise AI: the phase where boards want pace, CEOs want realism, employees want honesty, finance wants evidence, and the technology is strong enough to cause both value and mess. The companies that get this right will not be the ones with the most AI activity. They will be the ones with the clearest bridge between ambition and work the business can actually govern.

That is where AI becomes something the business can fund, govern, scale, and trust.

AI still needs translators and connectors. But the mandate has changed. The work is no longer to make AI sound understandable.

The work is to make it underwriteable.

◆

Stuart Winter-Tear is the author of UNHYPED and an independent advisor helping boards, CEOs, investors, and senior teams translate AI ambition into decisions they can fund, govern, scale, or stop.

For advisory enquiries: stuart@unhyped.pro