Selecting processes
Where AI can deliver an effect by the four criteria.
AI with a clear business task. We choose scenarios where AI changes the economics, check the data and prepare a pilot with a stop criterion.
Sound familiar?The team tries AI tools without a shared goal, It's unclear where automation is safe and useful, Data and processes aren't ready for implementation
AI and Technology
Where AI can deliver an effect by the four criteria.
Data quality, confidentiality, control requirements.
Tools and implementation scenarios.
One task, a baseline, success and stop criteria, 4–6 weeks.
AI changes the economics where a task repeats dozens of times a month, errors are cheap to spot, the result can be checked without an expert, and there's enough data. If a task meets three of the four criteria, a pilot is worth running. If it meets two, it's better to wait.
A list of scenarios scored against the four criteria.
A map of risks and controls.
A pilot plan with criteria.
The cost of waiting
Money will go to demos.
Employees will use AI haphazardly.
No one will know what data employees are already handing to outside services.
each has a metric and a target value.
whether they were named before the pilot.
whether it was named in advance.
"In which cases shouldn't this task be given to a model?" If the answer is "there are none," you're talking to a salesperson.
Besides development, budget for quality control: who checks the sample, who responds when quality drops, who updates the instructions.
In repetitive tasks, it reduces manual work. In tasks that need judgment, it shifts the work into review.
Services: AI Tools Implementation · AI in Workflows · Technology Architecture and AI Materials: AI in business: where it really changes the economics
A sales team that doesn't depend on one person
As long as results depend on one strong salesperson, you don't have a sales team. You have a person and a risk.