How we put AI to work at LOOKUP
I want our people spending more of their time with clients, solving problems and making decisions. There's plenty of preparation around that work which we can give them help with.
That's where we've put AI to work at LOOKUP.
We use it to research prospects, prepare for meetings, turn conversations into follow-ups and build proposals. We use it in our marketing, our business records and the way we keep track of commitments.
We started in our own business. I wanted to see where it helped, where it got things wrong and what our people needed to make it useful.
The biggest change has been writing down how we work and giving the AI something reliable to work from.
Give it the business knowledge
If I ask someone to prepare a proposal, they need to know what we sell, what we've agreed with the client, how we price it and what needs approval.
AI needs that information too.
It can produce a very convincing document without knowing any of it. Then you spend your afternoon correcting the prices, explaining the scope and rewriting it so it sounds like your business.
We've organised our business knowledge into plain documents. Our services, standards, processes and decisions have somewhere to live. Each fact has one home, so when something changes, we have one place to update it.
Before a job starts, the assistant has a brief. What we're trying to achieve, what information it should use, what a good result looks like and who needs to check it.
That information is useful to our people as well. A new team member needs to understand how we work. Someone covering a colleague needs the history. Writing it down helps both of them.
Develop it around the person doing the job
I think about this as a brain for the business, with an assistant developed around each person and their role.
The business knowledge is shared where appropriate. The help each person needs is different.
For someone in sales, it might be researching a prospect and getting the meeting preparation ready. For me, it might be bringing together a client's history before a conversation or finding the commitments that still need attention.
We call this The AI Executive Assistant Method.
You develop it much like you would develop a human assistant. Show it how you work. Explain what matters. Check the output. Correct it, then make sure that correction is available next time.
That last part matters. If you keep fixing the same thing without updating the instructions, you keep doing the same work.
Put it through real work
One of our working processes starts with a recorded client conversation.
We use the transcript to prepare the meeting summary, update the client record, draft the follow-up and capture actions with owners and dates. Those outputs use the account history and our written rules. I review the email and send it myself.
The meeting still needs a person. Listening properly, asking the next question and understanding what the client is worried about are part of my job. The assistant helps carry the preparation and follow-through.
We've also used it to prepare proposals from the information gathered during discovery and our service catalogue. The proposed scope and client commitments still need someone to approve them.
In marketing, we've taken an approved batch of social posts through scheduling and checked the saved posts against the plan. Approval happens before the software carries out that agreed work.
These are practical jobs with outputs we can inspect. We can check whether the record was updated, whether the draft reflects the conversation and whether the scheduled post matches what was approved.
Decide where people need to be involved
Our rule is simple: the assistant prepares, the person decides.
That means deciding the boundaries before giving it a job. Which records may it use? What may it change? What needs approval? Who handles an exception?
A written instruction is part of that. The access we give the software matters too. Telling an assistant to be careful doesn't replace controlling what it can reach.
We also use ordinary software for predictable work. A script can move information between agreed places or check whether a required field has been filled in. There's no benefit in asking AI to guess at something software can check directly.
I want the team to understand those differences. They need to know when to question an answer, how to check it and when to bring a person in.
When I'm coaching someone, I ask them to start with the goal. Tell the assistant what they're trying to achieve, give it the right information and check the result. Work through one layer, verify it, then continue.
Prove one before scaling it
We prove a process in our own workspace before transferring it to the team.
Start with one real job. Run it through. Check the output. Find what broke and improve the instructions. That gives us something concrete to teach and repeat.
The same thinking applies when we're looking at a client's business. We need to see how the work actually happens, including the exceptions, before deciding what should change.
Our Business Modernisation Framework puts discovery, security, modernisation, standardisation and optimisation ahead of the AI stages. That order helps establish a process worth automating and information we can safely use.
Measure whether it was worthwhile
Producing a document proves the document exists. It takes more to establish the benefit.
We need to account for preparation, checking, corrections and ongoing costs. We also need to understand what happens to the capacity we give back. Does someone spend more time with clients? Do proposals move sooner? Does follow-through improve?
For an implementation, I want a credible case for paying back the investment within 3 to 6 months. That's a test we apply when considering the work. The result still has to be measured after it's running.
The business knowledge and the process should remain useful as the tools change. Our people need to be able to read the instructions, improve them and carry them into the next system.
That's what we're building at LOOKUP. Help developed around the people doing the work, with the business keeping its knowledge and people keeping responsibility for the decisions.
If you're looking at AI for your business, bring us one job that keeps taking time away from your team. We can start by understanding how it works and where help would make a difference.