See the light!
empower my team with an agentic workflow

How I empower my team with an agentic workflow

When we decided to build with an agentic workflow, I was not worried about the technology. I was worried about the people. A new way of working only counts if everyone on the team can actually use it, and not only the engineers.

Here are the five decisions I made as a leader that I think mattered more than any tool choice.

1. One workflow for everyone, not a second-class one

The easy move is to give engineers the powerful process and hand everyone else a simplified one. I did the opposite. People handling roadmap, design and marketing use the same loop as the engineers: describe what you want in plain language, let the agent handle the mechanics, then check the result on a test environment.

The signal this sends matters. Nobody is on a “lite” version of the team’s real work. A decision made by someone who does not write code goes through the same path as one made by someone who does.

2. Opt in, never mandate

We ran the older way of working and the newer one side by side. Nobody was moved across on a deadline. People joined when they were ready, and the ones who waited lost nothing.

This was slower than a switch-over day, and it was worth it. A change people choose becomes a habit. A change people are told to adopt becomes a workaround.

On our marketing site build, the people in design, marketing and project management all chose to take part. They felt more capable and more efficient, because technical work no longer meant handing a request to an engineer and waiting for it to come back.

3. Put human attention where judgment matters

I kept people in exactly two places: deciding what gets built, and checking the real result before it goes further. The mechanical work in between, the build, the commit, the review, the deploy to a test environment, belongs to the agent.

That is the part of empowerment that is easy to miss. Giving people an agent is not the gift. Taking the process work off their plate, so their attention goes to the decisions only they can make, is the gift.

4. Make people owners of the outcome

Because the loop ends with a person verifying the live result, each person owns what they asked for from idea to working feature. They do not file a request and wait.

A teammate outside engineering planned and built lead source tracking for our advertising on the marketing site, spotted an issue in the existing setup along the way, resolved it, and then asked an engineer to verify the fix. The engineer became a reviewer instead of a bottleneck, and the work stayed with the person who owned the outcome. That is ownership and cross-function collaboration working together.

Ownership changes how people behave. They write clearer asks, because they will be the one checking the result, and they care about the outcome, because it is theirs.

5. Know the foundations well enough to direct, then give the time back

This one is about me. Directing agents well worked because I already knew our stack deeply. Depth is what lets a leader give clear direction and spot a wrong turn early. If you are going to empower a team with agents, expect to invest in knowing the foundations yourself first.

The payoff is what I value most: time. There is less manual coding and troubleshooting, so my day moves toward architecture and thinking, and toward removing blockers for the team.

For the team, the saved time went into thinking. It gave each person room to think, design and create more value for our customers, for each other and for the company.

What I would tell another leader

None of this is about the agent being clever. It is about what you ask your people to spend their attention on. Make the workflow the same for everyone, let people join by choice, keep humans on the decisions and the checks, let them own results, and know your foundations well enough to lead.

We built and launched an AI search product this way with a lean team, and wrote up the journey in our case study. For the broader picture of what the search product does, see our guide to AI search.

We built our client support portal the same way, and we are building more products with this workflow.

If you look after a website, our client support portal gives you a real-time view of its activity, performance and support history.

Disclosure: an AI agent drafted this post and I reviewed, edited and published it. I am a tech enthusiast at WhooshPro, which makes the products I mention.