We talk to you and the people who know the business. We examine actual enquiries, orders and exceptions with the teams involved. Where the decision depends on demand or the customer experience, we include customers or dealers too.
That can reveal a gap between what you want to offer and what the team can currently deliver. Better access to product information may be enough. Sometimes responsibilities, capacity or the offer itself need to change. We assess the software alongside those choices.
We apply the same judgement to AI. What expertise does an employee need to move forward? Would better access to information or conventional automation be enough? Could AI help assess a request while leaving the decision with the employee? An agent becomes relevant when the work requires next steps that depend on intermediate findings. We assess whether the information and permissions are available, failures can be recovered from and the value justifies the cost.
For AI, we prefer an open model in a private environment, on a virtual private server (VPS) where appropriate. We favour European models when they meet the needs of the task. Model quality, processing location, access and running costs are assessed together: a larger model needs to justify its extra cost. Our approach to private AI and targeted use of agents explains those decisions.
We review the plan with the people who will put it into practice. Can a dealer prepare a sound quotation? Can production assess an exception? Can someone see who needs to decide next? Those situations tell us whether the plan works.