A year into the Operating Director role with Phoenix, self-confessed ‘second quartile tech adopter’ Dan McManus unpacks how business leaders can master AI to deliver cohesive and effective people strategies, while keeping the human factor at the centre of the process.
My calibration problem
I’m a reliable second quartile adopter of technologies. It took me until 2017 to trust myself to book a flight on my phone, and 2019 to get a smartwatch, so trying an LLM in early 2024 was fairly typical behaviour for me.
What happened next, however, was a little more unusual. I moved quickly from creating novelty poems (remember those!) to putting every email through a rewrite, and then I stopped. I came back to it later, and started outsourcing more content creation and synthesis.
Before long, I realised I was outsourcing the kind of difficult and high-friction thinking that was useful precisely because it was difficult and high-friction. So I changed how I did things. Over the last year, I gradually found something closer to a genuine working rhythm: one where LLMs amplify my effectiveness without dehumanising my contribution.
Much of that calibration depended on me getting a clear sense of what I actually do, and what others value about how I do it. That challenge goes up to the organisational level. Leaders are grappling with AI and its implications for their working life, while also designing the environment in which everyone around them is doing the exact same thing.
So how will AI impact the world over the next decade? On the scale of “Google on steroids” at 1 to “civilisation-scale inflection point” at 10, my money’s on an 8. Much of the value will come from models embedded deep in workflows, products, or proprietary data, and there is plenty written about that.
What interests me is this: no matter how far they have travelled on their AI adoption journey, every leadership team must question how their people use, question, learn from, and stay accountable for AI-enabled work.
The businesses that handle this best will not necessarily be those that treat the speed at which they can strip out people cost and ramp up token usage as the twin metrics of their success. They will be those whose people strategy is oriented around building a better, AI-enabled, appropriately resourced business.
Five things worth getting right
1. What leaders are actually responsible for
AI is very good at making things look convincing. A well-prompted model will return something structured, fluent and plausible. That veneer of credibility is precisely the issue.
The question a leader should be asking of any AI-assisted output is not whether it looks right, or even whether it was AI-generated. It is whether the reasoning is sound, whether the right question was asked in the first place, and whether there is a human who owns the conclusion and can defend it.
Accountability can fall by the wayside when everyone assumes the machine checked the thinking. Leaders need to understand that AI is not here to make key strategic decisions for them. Instead, AI tools represent an evolution in the entire leadership role. It exists alongside the established workflow, and like any part of a process, it must be held to scrutiny. Leaders should exercise a little scepticism when receiving and stress-testing ideas that look impressive on first reading, and show greater discipline in ensuring that accountability for outcomes sits with a person, not a process.
2. Org design that starts with questions and incorporates evidence
The temptation in most businesses will be to let org design follow the path of least resistance: wait until AI has demonstrably changed the volume or nature of repeatable tasks, then adjust headcount accordingly. That is a strategy of sorts, but it is a reactive one, and it tends to optimise for cost rather than capability.
A more deliberate approach starts earlier, identifying specific processes where AI changes what is possible rather than just what is affordable. Not abstract functions, but contained, meaningful workflows that span the breadth of a business: graduate sourcing, cash collection, first-pass contract review.
The question for each is not just “can AI do more of this?” but “if AI does more of this, what does the human role become, and is that a role worth designing for?” Some answers will point toward reinvention. Others will honestly point toward reduction. A coherent people strategy requires the intellectual honesty to distinguish between the two before events make the decision for you.
3. Onboarding that addresses both AI ‘cheating’ anxiety whilst allowing space for human learning
New joiners are entering organisations with highly variable personal relationships with AI. Most onboarding processes have not caught up with that reality. The risk cuts in two directions, and both of them matter.
The first is anxiety-driven avoidance. Without a clear organisational signal that using AI is not only acceptable but expected, many new joiners will simply avoid it, worrying that a manager will read AI-assisted work as evidence they are not pulling their weight. For these people, it is the workplace equivalent of smuggling a calculator into a mental maths test: the tool would help, but using it feels like an admission of something. That equivalence is a failure of culture and communication, and it leaves capable people working less effectively than they could.
The second and opposite failure is a new joiner who runs every draft through AI, who never digests and thinks through feedback themselves because there’s always an LLM that can implement it more quickly. They may look productive earlier than they actually are, but they are skipping the friction that builds judgement. An onboarding process with a coherent view on AI should be explicit about where that friction is valuable and worth preserving, and where AI genuinely accelerates learning rather than replacing it.
4. Performance frameworks that reward judgement, not polish
Performance assessment has always been an imperfect science, but AI introduces a specific new distortion. When a tool can generate polished, plausible, well-structured output at speed, the traditional proxies for strong individual performance become less reliable signals of genuine contribution.
Volume, presentation quality, apparent thoroughness: these are no longer safe measures of what someone actually brings. In an AI-assisted environment, fluency becomes cheaper. Judgement does not. A business that doesn’t update its frameworks will find itself rewarding the wrong things, and over time, developing the wrong capabilities.
The shift required is from measuring what people produce to measuring their impact on outcomes. That means assessing the quality of the questions someone asks, the judgement they apply when AI gives them something credible but wrong, and their ability to turn neat AI-generated output into real change in the world. It also means treating structured, intentional AI use as a positive signal in itself: evidence of someone who understands their tools and deploys them well, rather than something performance frameworks ignore or implicitly penalise.
5. Retention and succession: inaction isn’t protection
Some roles will disappear as AI adoption becomes more widespread and deeply integrated into modern workflows. The direction of travel is now beyond all doubt, and a people strategy that pretends otherwise will lack credibility with the people it is supposed to serve, whilst ultimately making a business uncompetitive.
That reality makes the case for moving thoughtfully but deliberately. A business that drags its feet on AI to avoid difficult conversations about roles isn’t protecting its people. It is simply deferring the disruption while competitors build capabilities it doesn’t have. The people inside that business are not being shielded; they are being developed for a version of their role that has a shorter future than it might otherwise have had.
Retention in this environment is therefore less about reassurance and more about demonstrating a credible path for the individual and the organisation. People will stay, and commit, where they can see that AI adoption is genuinely oriented around making them more capable and more valuable, not just making the business leaner.
Succession planning needs to keep pace with that too. A pipeline built on developing people for roles as they exist today, rather than as they are becoming, is a pipeline that will disappoint.
Acting with intent
Every business will find its own version of this challenge. But those that navigate it best are likely to be the ones already asking hard questions at every level: how leaders interrogate and own the outputs they receive, how organisations are designed around genuine transformation rather than reactive cost reduction, and how individuals are given the clarity and confidence to work with AI in a way that makes them more capable rather than less human.
The technology will keep moving. A coherent and deliberate people strategy is what ensures the people move with it, rather than being moved by it.
