The problem
Every AI programme starts with a long list. Operations, servicing, finance and IT each have a workflow they want automated, and someone has to decide which one goes first. In most organisations that decision is made in a workshop, from whoever argues best, and written up in a slide that nobody can check later.
The work behind a good decision is tedious. You interview the people who do the job, read the process documents, pull volumes and handling times from the ticketing or case system, and estimate what an agent could realistically take off their hands. Then you have to compare workflows that are measured in different units: cases per day in one team, hours per month-end close in another. The estimates end up in a spreadsheet with hard-coded numbers, and by the time the steering committee sees it, nobody remembers which figure came from a system export and which came from a guess in an interview.
Good looks like this: a short ranked list, each workflow with a payback range rather than a single confident number, the one metric you would use as its baseline and where you would read it from, and a clear list of what is assumed versus measured. A sponsor should be able to point at any number and ask “where did this come from?” and get an answer.
How it works
Forge Scout is a Blueprint. We build it on demand in your engagement, usually in the first week of an assessment. This section describes the approach we would take.
It reads three kinds of input for each candidate workflow: notes from interviews or workshops, the process documents or standard operating procedures, and exports of volume and effort data as CSV. Every line and every row is numbered, so each figure it uses can point back to its source.
It is steered by a scoring model that you agree with us before the run: what counts as value (hours saved, cost per case, cycle time, error rate), what counts as feasibility (data access, how structured the inputs are, regulatory constraints, how often a person must stay in the loop), and how the two are weighted. The scoring model is written down and versioned, so the ranking can be rerun when an input changes.
It writes one card per workflow: the current volume and effort as found in the sources, the share an agent could plausibly handle as a range, the resulting payback range, a proposed baseline metric with where to measure it, and the risks that would sink it. It then ranks the cards. Anything it had to assume is listed separately and never mixed into the measured figures.
Then it is checked. Every number on a card has to cite an input line or a calculation from cited numbers; a figure with no source is rejected rather than shown. That is a structural check, not a judgement that the estimate is right.
People decide. Your business owner and our engagement lead review the ranking, challenge the assumptions, and sign off the baseline metric for the workflow chosen. Scout prepares the decision; it does not make it.
Where it gets it wrong
The ranking is only as good as what goes in, and the failure modes are predictable.
- Interview numbers are optimistic. People overstate how long a task takes when it annoys them, and understate it when it is their expertise. Scout marks interview figures as reported, not measured, and the card says which is which. Where a system export exists, it wins.
- It cannot see the work nobody wrote down. Exception handling, workarounds and the spreadsheet a team keeps on the side rarely appear in process documents. Scout raises missing steps as open questions; a short walkthrough with the team closes them.
- Feasibility is easy to misjudge from documents. A workflow can look ideal on paper and then turn out to need data the team cannot reach, or approval that will take a quarter. Scout lists data access and approvals as risks to confirm, and the scoring model can weight them heavily.
- Payback assumes adoption. A workflow that saves hours only pays back if people use the result. Scout states the adoption assumption on every card, so the sponsor can see how sensitive the ranking is to it.
We would rather Scout show a wide range with honest assumptions than a single neat number that falls apart in the first steering meeting.
Why not just prompt a chatbot
A chatbot will happily rank your workflows and produce a confident ROI table. You will not be able to tell which numbers came from your data and which were invented to make the table look complete, and the ranking will change if you ask again. What makes Scout useful is the agreed scoring model, the rule that every figure cites its source, the separation of measured from assumed, and a ranking you can rerun when an input changes. That is the part we build with you, and the part you keep.
What you get
- In your engagement: built on demand in your engagement, and used in the assessment week to rank your candidate workflows and agree the baseline before anything is built.
- Under your brand: services firms can run it inside their own assessment offer, with their own scoring model and report format.
- Built to own: we tune it to your value and feasibility criteria and hand over the source, so your teams can rank the next round themselves.