From senior to lead (Part 2 of 2)

We ended off Part 1 with a real-world example from yours truly, and as a recap, it was all about managing up. That article got way too long and I decided to split it into two. Part 1 covered doing the right things and managing up; this one covers doing them the right way and managing in all directions, with a dose of reality from a post-AI world where execution has gotten cheaper.

No one is checking in on you

Taking the lead means you’re accountable for everything, and it’s a lot like running your own business. Once you’ve set your own OKRs and your boss has agreed, the deadlines and commitments are yours to promise. No one is outlining the work to be done for you, chopping it up into release plans and giving you a schedule.

We’ve already seen how much of leading work is communicating and talking to people. The enabler is your reputation. You want people to not just give you airtime to speak, but to commit to things and rely on your work. And so, a large part of being a leader is growing and maintaining that reputation, without which you can’t really achieve much.

Managing risk

When promising outcomes in the future, you’re making a bet and taking on risk. Nothing is ever 100% certain, and I promise your boss already understands this keenly. Your failure is also their failure. The good thing is, managing risk is in both of your best interests, which means you are already aligned on something.

The best thing you can do for yourself is installing feedback loops that help you find out if you’re wrong quickly. State your bet upfront clearly so you know what wrong looks like. Set up cadences and systems of accountability with your peers or your one downs. Hopefully, you have been nurturing their growth so that they’re actually capable of supporting you in this sort of ambiguous work.

Keep tabs on moving parts that you don’t have control over, and check in often. The slightest whiff of derailment should set off your alarm bells to step in and course-correct. A direct report who isn’t communicative and holes up to finish their work is a risk. Set up systems of accountability not just for yourself, but also for work that you coordinate, both inside and outside your team.

Communicate early and often

In the event that you do hit a dead end, you’ll realise it sooner than your boss. Always seek feedback early, and communicate changes early. Don’t wait till you can schedule a meeting; a short message or email to let your boss know what changed is incredibly valuable—they have their own timeline and ongoing conversations with other leaders. You don’t want them to make verbal commitments not knowing the latest developments that only you have visibility into.

Stale information is more harmful than wrong information. Outdated things were correct at one point in time, and they remain believed correct until new information arrives. They can also remain sticky in your company’s corpus far longer than you’d like to: text messages in a Teams chat, unmaintained Confluence pages, video recordings, fine-tuned LLMs and data stored in vector databases for RAG… and even humans repeating them in meetings. The worst part is that stale information cascades into parallel workstreams, which in a large org is a nightmare to stop.

It’s still all about people

Fundamentally, you’re not just making a decision for yourself, you’re also making it for other people. You’re making decisions on behalf of your very own teammates, collaborators who rely on your work, and other leaders who rely on your promises. This already makes it a high-stakes decision, because one wrong turn and you’re burning precious resources at least twice—once for doing something with poor returns, and once more for steering the team back to the right thing. You could also be burning bridges.

As if it wasn’t hard enough to convince smart people to follow your lead, you also have to be careful of borrowed conviction flowing downwards. Just as you nod your head in agreement with your boss, your team will do the same to you. How do you tell buy-in from deference? Different strokes for different folks. Some people push back as a way to test your idea, while others propose alternatives as a way to softly disagree. What you don’t want is for people to wait for instructions and fully agree in meetings. It’s usually a warning sign if there are no questions raised.

Delegation

After putting in all the legwork to have a point of view and communicating that vision, you don’t end up doing the work yourself. You physically can’t because there’s too much. Sorry, it doesn’t get better here for ‘craftspeople’. What was hard for them to relinquish control of remains the same in a post-AI world, whether you’re delegating to humans or machines. Doing things the right way as a leader means letting go and creating the system and environment for others to do so.

I’ve reflected on this same thing in an article I wrote ~6 years ago¹, and it still holds true today. Design yourself out of the way. One measure of your success in delegation is that the work gets completed without you being in the room. I’m not about to expand into how you can do this, as it would double the length of this article. Smarter people have spilt plenty of ink on this topic already in best-selling self-help and leadership books. However, I’ll still touch on what I think is one of the more important things to get right: communicate early and often.

Yes, we’re doing the same thing managing upwards and downwards. As much as you communicate early and often to your superiors, your direct reports relying on you for context also need frequent updates so they can course correct quickly. You cannot afford to hole yourself up.

In fact, all of the ideas discussed so far in being a lead apply upwards, sideways and downwards. We’re using the same skills in all directions, because the work is really all about people, and people are constant moving parts.

Leading in a post-AI world

In the opening I’ve said that execution has gotten cheaper. That’s actually only half the picture, because it’s only going to get cheaper from here. In such a world, bottlenecks move to verification and coordination work, and what has become even more scarce is the skill in choosing the right things to do.

In the design world, the current conversation revolves around having ’taste’ because AI can do most of the technical work now. It’s the same in any other domain. LLMs perform exceptionally well with good context, and the most valuable context is a point of view. Guess where that comes from?

You.

Even more borrowed conviction

One big danger I see is how AI usage affects your cognitive habits subconsciously. This is a throwback to ‘borrowed conviction’ in Part 1. Anthropic recently published research on ‘How AI agents shape experts’ judgment calls, and the findings were not great:

… analysts went with their agent’s lean 86% of the time. If the agent’s input made no difference, we’d expect about 50%. When analysts had already expressed their own preference before the agent weighed in, they switched away from it 60% of the time if the agent disagreed, compared to 6% if it agreed.

Dig into it and read the whole thing for your own sake, and know that this affects everyone. Not only do you have to guard against borrowed conviction from your team members, you also have to guard against yourself. Remember that ‘doing the right thing’ is about having your own point of view—make sure that it’s your voice and not a machine’s.

Conclusion

We’ve talked a ton about what a leader should do, mostly cerebral things, or the mind of a leader if you will. I’d like to end off with the heart of the matter.

As a leader, you lead people. The day will come when you no longer lead them, and you’ll know if you did a good job when they decide to follow you even when they don’t have to.