Baseline

In one paragraph

A baseline is a record of how a piece of work runs before anything changes. It captures how long a task takes, what it costs, and how good the result is, measured the ordinary way people already work. Once an AI system is introduced, the same measurements are taken again and compared against it. Without a baseline, a result is an impression. With one, it is a number that can be checked.

Why it matters

Most organisations can point to an AI system people like using. Far fewer can say what it actually changed, because nobody wrote down how the work ran beforehand. By the time someone asks for evidence, the old way of doing things has often stopped, or people cannot recall it clearly enough to be sure. A baseline fixes that starting point in writing, so a later comparison rests on something solid rather than memory.

How it works

A baseline records three things for a unit of work you can repeat: a case handled, a document processed, or a decision made.

• Time: how long the task takes from start to finish, measured consistently.
• Cost: what goes into completing it, counted in full rather than as a single line item.
• Quality: whether the result was right, complete, and fit for its purpose, judged against a fixed standard.

These three are measured the same way before the AI system is introduced and after, so the comparison is fair. If your organisation started using an AI tool before anyone measured how the work ran beforehand, the baseline can still be captured while part of the work is done the old way in parallel. Once set, the baseline becomes the yardstick you keep returning to, not a one-off exercise: it is what later checks are measured against as the system changes, gets tuned, or is replaced.

What it looks like in practice

Picture the work before anything changes. A team handles a steady stream of customer requests, and before an assistant arrives to help draft replies, someone records how long a typical request takes to close, how many need to be reopened, and how a sample of replies score against a simple quality checklist. Once the assistant is in use, the same three measurements are taken on the same terms. The difference between the two sets of numbers, not the number of people using the assistant, is what tells the team whether it helped.

How this connects to our work

Establishing a baseline is the first step in measuring what an AI system costs and delivers, and in showing that an automated workflow genuinely shortened the work. Every measurement we run starts by capturing how the work is done today, because a result without that starting point cannot be checked against anything.