Start with the result you need
A vacancy often starts with the description of the person who previously held the role. That is a useful reference, but it can preserve tasks that no longer need to be organised in the same way. Before reopening it, describe what customers and colleagues need from the role. Make that result concrete enough to discuss with the people doing the work.
For example, imagine a service business where someone prepares customer requests for a specialist. The useful result is a complete, accurate brief that lets the specialist respond. Collecting documents, finding earlier decisions and asking for missing information are steps towards it. This is an illustrative example, not a claim about a customer implementation.
Look at one responsibility rather than redesigning the entire organisation at once. Ask what good work looks like, which decisions require expertise and where people currently spend time repeating or reconstructing information. Those answers give management a better starting point than a list of AI tools.
Keep judgement visible
An agent can help organise information, prepare a draft or flag gaps. Someone still needs to know whether the result makes sense in the customer’s situation. A convincing answer and a responsible commitment are different things. Design the role so that this distinction remains clear.
In the service example, the agent might assemble an initial brief from approved sources. The responsible person checks unusual requests, resolves conflicting information and decides when the specialist must be involved. The company determines who may promise a delivery date or approve a change in scope.
This preserves the value of specialist knowledge. A broader view of the work should help people recognise when they need an expert. It should not encourage them to act as experts in every field. Hire and coach for curiosity, sound judgement and the ability to ask a precise question, alongside the knowledge the role genuinely requires.
Design the role with the people doing it
Ask employees to walk through an actual request from beginning to end. What do they receive? What do they have to find? Which missing detail causes a return trip? What does the next colleague need to continue? These conversations make an abstract discussion about AI much more useful.
A practical exercise is to spend time with the adjacent team. A commercial colleague can learn which promises complicate delivery. An operations colleague can learn what the customer was trying to achieve. The purpose is understanding, not expecting everyone to master every profession.
Then agree what changes for the person. If agents take over preparation, where should the recovered attention go? It might go to a better customer conversation, resolving exceptions or improving the service. A wider role needs clear priorities, suitable authority and a manageable workload. Simply adding responsibilities will not create a better organisation.
Build the support around the responsibility
Once the role is clear, build the agent support to match. Define the information it may use, the preparation it should produce and the point where someone takes a decision. Give current instructions an owner so the system can follow the way the company actually works.
For the illustrative request workflow, start with one request type and a small set of approved information. Agree how the agent marks missing facts and how the responsible colleague corrects the brief. Preserve the corrections that should improve future work, with appropriate access to customer information.
This is where strategy and hands-on building belong together. Management decides the responsibilities and boundaries. The implementation makes those choices usable. Coaching helps people direct the agents, challenge weak outputs and recognise exceptions. A diagram alone does not complete that change.
Test the role before rewriting every vacancy
Choose a bounded trial and compare the whole task with the current process. Measure preparation time, corrections, missing information and the time the specialist needs to reach a useful answer. Include the effort spent supervising the agent. Faster drafting is only helpful if the work that follows becomes easier too.
Talk to the people involved after the trial. Did the brief help? Were decision rights clear? Did checking the output consume the time it was supposed to save? Did the customer receive a clearer response? Use the answers to adjust the role and system before expanding.
The next hiring decision can then reflect observed needs. You may need deeper expertise, stronger customer judgement, better coordination or additional capacity. Agents do not determine that decision for you. They give you another way to organise the work, which management must evaluate responsibly.
Build a company that can use the capacity
My approach is to help define the business outcome, shape the responsibilities with your people and build the agent support behind them. That can serve a new business or an organisation improving the way it already delivers to customers.
Start with one responsibility your team would like to improve. Describe the result, name the decisions that matter and identify the repetitive preparation around them. That is enough to begin a useful conversation about what to build and what kind of person should lead the work.