The short answer.
After the first AI workflow comes the workflow that receives its output, or the same piece of work in the next team. You choose that second workflow from your plan, not from whoever is most excited. Once the first workflow is in daily use, you record what you learned, pass the method to a second owner, and start one more workflow per team, with no more than two or three in development across the company at once.
The plan gains one new section to hold all of this: a “now, next, later” list that the plan’s owner reviews every month. That list is what stops the first workflow from being the only one.
The second workflow continues the first.
The best second workflow picks up where the first workflow stops. There are two ways to find it.
- Follow the output downstream.The first workflow produces something: a brief, a packet, a checked list, a draft. Someone receives that output and does the next piece of work with it. A workflow for that next piece of work starts from records already clean and reviewed, so it is quicker to develop and test.
- Take the same work to the next team.Most companies repeat a piece of work in more than one team: a weekly status report, a client handover, a month-end checklist. The second team adapts a method that already passes its tests, and the first workflow’s owner already knows where the hard cases are.
Both routes build on what the company already has. How to deploy AI workflows to a whole team, and keep them working covers the rollout of the first one before you widen it.
Choose the second and third from the plan.
The second and third AI workflows come from the plan you already wrote, scored the same way you scored the first. After a first workflow works, every team asks for one, and the loudest request is rarely the right one.
Score the candidates you set aside the first time again, with what you now know. Choose your first AI wins has the scorecard. Four questions settle the order.
- Does the candidate receive the first workflow’s output?If yes, it moves up. Its inputs are already in shape.
- Does the same work happen in another team?If yes, it moves up. The method already exists.
- Is there one person, by name, who will own the candidate?If nobody steps forward, the candidate waits, however good it looks.
- Can a person check every output before it leaves the team?If not, the candidate waits until there is a review step.
The plan decides, and the plan’s owner says so out loud. A team that hears “you are next, in this order” waits happily. A team that hears nothing starts its own experiment, and the company drifts back to undirected use. From a few AI pilots to the whole company covers that drift and how to stop it.
Three things to do once the first workflow is in daily use.
A first AI workflow is in daily use when the people who own the work run it without being asked. At that point, do three things before anything new starts.
- Record what you learned about the work.Writing the first workflow exposed things nobody had said aloud: a check that only one person did, a step two people did differently, an input that was often missing. Write those findings into the procedure itself.
- Pass the method to a second owner.The person who owned the first workflow now knows how to capture a piece of work, test it on real cases and roll it out. Have that person sit beside the owner of the second workflow through its development. The company stops depending on one enthusiast.
- Start one more per team, two or three in total.Each team with a workflow in daily use starts its next one. Across the company, keep no more than two or three workflows in development at once. Each one needs its owner’s attention while it is tested on real work, and attention runs out before ideas do.
The first workflow still needs its owner after the second one starts. Keep it working: a workflow in its sixth month sets out the owner’s routine, and Keep it working is our offer for the months after handover.
A now, next, later sheet to copy.
The “now, next, later” sheet is one page added to your AI plan. Copy these headings into a document, fill them in with the leads of each team, and review the sheet every month.
- Now: in daily use.Each workflow the team runs every week, its owner, and the date it went into use.
- Now: in development.No more than two or three across the company. For each one: the team, the owner, what it starts from, what it produces, and who decides.
- Next.One workflow per team, in order. For each one: why it is next (receives an output, or repeats in another team) and the named person who will own it.
- Later.Every other candidate, with one line on what it is waiting for: an owner, a review step, clean records, or the workflow ahead of it.
- Left alone.Work you decided AI will not prepare, and why.
- What we learned this month.Findings about the work itself, written back into the procedure.
- The monthly review.Who reviews the sheet, on which day, and the three questions: is each workflow in use, is it useful, and what moves from next to now.
Illustrative example.
Take a 120-person property management company. Its first workflow, owned by the leasing manager, turns each rental application into a checked file for the person who approves tenants. Once the leasing team ran it daily, the operations director went back to the plan. The work that receives the leasing team’s output is the move-in: the maintenance team inspects the unit and the accounts team sets up the tenant’s billing. So the move-in packet went to “next”, owned by the maintenance coordinator. The same application check also happens in the commercial leasing team, so the commercial version went to “next” as well, owned by the commercial leasing lead. The leasing manager sat in on both. Marketing’s request for listing descriptions went to “later”, waiting for an owner. The operations director reviews the sheet with the team leads on the first Tuesday of each month, and nothing moves from “later” to “now” without passing through “next”.
What is different in a company of under 200 people.
A company of under 200 people expands from one AI workflow to many with less ceremony than a large enterprise. Most published advice assumes a central team, a platform decision and a program office. You need none of them.
- The owner of the plan knows every team lead.The order of work is settled in one conversation, and every team hears the same answer.
- The second owner sits down the hall from the first.The method passes from person to person, which is how it spreads best.
- Downstream work is easy to see.In a smaller company one person can name who receives each output, so the second workflow is obvious.
The constraint is attention, not budget. A small company runs out of people with time to own a workflow long before it runs out of ideas, which is why the limit of two or three in development matters more here than anywhere.
Keep going, one workflow at a time.
The second workflow is where a company moves from one good experiment to using AI on purpose. How organizations plan their use of AI sets out the whole method, and We have a plan. How do we build the first AI workflow? covers the first build.
FullyPowered Planner is where the plan and its now, next, later list live. You set out your teams, their roles and the work between them, and choose what to build next. Colleagues add their own teams. When you want help with the next one, we develop it with you in a 3-Day Workflow Sprint or a 2-Week Workflow Sprint, starting from a free template in the Workflows Directory where one fits.