Leaders need to figure out their AI stack

Leaders who think AI is just another tool they can plug into existing workflows are getting this wrong. This is not a software upgrade. It is a workflow redesign moment.

Every function head now has to figure out their AI stack. They also have to figure out what more becomes possible once that stack exists.

Earlier, a leader’s operating stack was a set of tools their team used to get work done. LinkedIn. ATS. CRM. Figma. Email. Calendar. Docs. Now the stack is shifting from tools people use to agents that do parts of the work.

That changes the job of leadership. Leadership has always been about setting direction, building teams, and creating the conditions for good execution. Now it also requires redesigning how the function works when agents can take over chunks of execution, and rethinking the scope of what the team can actually do.

If you do not do that redesign yourself, your team will keep using AI as a thin layer on top of an old system. You will get marginal gains, not a new operating model. You will also miss the chance to expand the ambition of the function itself.

Take talent acquisition.

Earlier, the workflow looked roughly like this: scout LinkedIn, identify candidates, reach out, scan resumes, do an initial screening call, coordinate interviews, and move people through the pipeline. The operating stack was LinkedIn, an ATS, phone calls, email, and Google Calendar.

Now imagine that same workflow rebuilt for an agentic world. One agent scouts LinkedIn against the role brief. One agent scans resumes and ranks candidates based on the traits the team actually cares about. One agent runs the first round to collect structured signals. One agent coordinates calendars with interviewers.

The TA lead is no longer just managing recruiters using tools. The TA lead has to design the system of agents, handoffs, review points, and escalation rules that make hiring work.

That is the new stack.

And this does not stop with hiring.

Take sales.

A sales leader cannot treat AI as a note taker added to the current process and call it transformation. They have to ask bigger questions.

Which parts of prospecting should agents do? Can an agent research accounts, generate first-pass outreach, update the CRM, prepare call briefs, summarize objections, and suggest follow-ups? Where should a rep step in?

What signals need human judgment? What should be fully automated, and what should never be?

The new sales stack is not just Salesforce plus a few AI copilots. It is a set of agentic workflows across prospecting, qualification, follow-up, forecasting, and pipeline hygiene. That stack needs an owner. That owner is the sales leader.

And once that stack is working, the question changes from how to save rep time to how much more pipeline the team can cover, how much better prepared every conversation can be, and how much tighter execution can become.

Take design.

A design head also cannot think in terms of just adding one more AI tool to Figma. The real question is how the workflow changes.

Can agents turn product requirements into rough flows? Can they generate first-pass explorations for multiple directions? Can they evaluate consistency against an existing design system? Can they help turn user research into recurring patterns and insights? Can they help designers move faster from mockups into code that gets pushed directly into product?

If so, then the design leader has to decide where human taste, product sense, and judgment matter most. The job becomes designing a workflow where agents expand the team’s range without flattening the quality of thinking.

This is bigger than adding another tool to the design stack. It is about rethinking how design work gets done, and how much more the team can ship.

It is also a scope question. Once designers can move from idea to shipped product faster, the team can take on more experiments, iterate more aggressively, and operate closer to the product surface than before.

This is why I think every function head needs to figure out their AI stack now. Not later. Not after the market settles. Not once some central innovation team publishes a playbook.

Right now.

Because the people who redesign workflows early will learn faster what humans should own, what agents should own, and where the real leverage sits. They will also learn what new output, speed, and range become possible once the work is rebuilt around that stack.

Everyone else will keep layering AI onto legacy processes and wondering why the gains feel underwhelming.

The uncomfortable truth is this: AI is not just changing execution. It is changing management. Leaders now have to build the new operating stack for their teams.

If you lead a function, that is your job.

You do not need a perfect answer on day one. But you do need to start mapping your workflows, breaking them into steps, identifying which steps can be agentic, deciding what the new stack should look like, and asking what your team could do once those constraints start to fall away.

Because in this phase, the winners will not just be teams with access to AI tools. They will be teams whose leaders figured out how work itself needs to change, and what new scale of work becomes possible because of it.

Hope you enjoyed the read! If you have feedback or a different perspective, I’d love to know. Catch me on Twitter or mail me at me@chettyarun.com Thanks!

The thoughts are Chetty Arun’s, but he used Hoid - his blog writing agent - to shape and publish this post.