<!-- Markdown-versie van https://outdeliver.ai/en/blog/which-process-first, gemaakt door tools/bouw.mjs -->

Blog

# Which process deserves AI first?

Do not start with the biggest problem. Pick a frequent task with clear data, a
safe boundary and an owner who can measure it.

The first question is rarely whether AI can technically do something. For many
tasks the answer is at least partly yes. The more useful question is where a
first change proves itself quickly enough to build on. Not every frustrating
process is suitable, and nor is every high-volume one.

A good first choice recurs in a recognisable form and has a clear start and end,
data that is available and an employee who knows when the outcome is right.
Think of an order email that has to go into the ERP system, an invoice that
needs checking or a customer question that needs information from a fixed set
of sources. None of these are spectacular. That is exactly why they can deliver
value without overhauling the whole company at once.

## Put the candidates side by side

Pick three to five processes that staff currently describe as time-consuming.
For each one, ask how much frustration it causes, and also how many cases there
are per week, how long an average case takes and how many exceptions occur.
Note where the data comes from and which system receives the outcome.

Then ask six questions. Does the task come up often? Is the starting point
recognisable, such as a mailbox, form or queue? Are the steps mostly clear? Is
the source data usable? Can an error be put right? And does the process have an
owner who can decide on exceptions, measurements and changes?

Add one hard exclusion rule: do not start without a process owner, a recovery
route and confirmed access to the data you need.

A candidate with high frequency and clear steps is usually a better start than
a rare, complicated decision. A frequent but complicated process need not be
ruled out. Split it up first. For example, let only the recognisable order
lines from regular customers go through, and keep unusual prices or new
customers in a queue.

## Use the matrix to start a discussion, not to keep score

A simple matrix has two axes: frequency and complexity. Frequent and simple is a
candidate, provided errors can be put right and a process owner can measure the
outcome. Rare and complex is seldom a sensible start, even when the demo looks
impressive. The other two boxes call for a choice. Frequent and complex usually
holds a lot of value, but has to be broken into smaller steps first. Rare and
simple can wait, because the time saved stays small.

The matrix is not a calculation that makes the choice for you. A task can be
frequent and still unsuitable when the source data is very poor. A task can be
rare and still urgent because of a safety risk. That is why the process owner
belongs at the table. They know which exception hurts, which rule already exists
and which results matter to the organisation.

## Make the first scope smaller than the name of the process

"Order processing" is too big. "Standard order emails from regular customers,
with no price discrepancies, turned into draft orders in the ERP system" is a
first scope. The second description shows straight away what falls outside the
first version. The measures follow from it too: how many of these messages go
through without correction, how long they take and why some go to an employee.

Also decide in advance when you will stop. After a week it may turn out that
the source emails vary too much, that the volume is lower than expected or that
the current way of working is already cheap enough. Stopping keeps you from
polishing an unsuitable process for nothing and leaves room to pick the next
candidate.

## Build a foundation for the next step

A first project has succeeded when the process is more than a demo and goes into
use with an owner, checks and a measurable outcome. Then it delivers more than
time saved in one inbox. The organisation learns how to connect sources, handle
exceptions, grant permissions and test changes. All of that carries over to the
next process, so you do not have to start from scratch every time.

So pick the task that comes up often enough, can be scoped safely and shows
straight away whether the new way of working really improves on the old one,
even if it is not the most impressive.

Three ways to start

## Want to compare three candidate processes side by side?

Start small, with one process that comes up every week. Choose how:
from half an hour behind the scenes to a session at your offices.

### Look behind the scenes

Free

Half an hour, by invitation

See everything we build running in our sandbox, the Proeftuin. No sales pitch, unless you ask for one.

[Request an invitation](https://outdeliver.ai/en/microsoft.html#toegang)

02 Recommended

### Introductory call

Free

Half an hour on Teams

You tell us what you already do with AI. We tell you what fits, or whether it is still too early.

Call with Marcel Logtenberg

[Book a call](https://outdeliver.ai/en/book-a-call.html)

### Get started

From €495

At your offices, excluding VAT

An AI quick scan, an AI assessment or an AI assessment plus inspiration session, at a fixed price.

[Get started](https://outdeliver.ai/en/get-started.html)
