An agency decides to offer assistants across its client base. The instinct is to propose it to everybody, since it could help anybody.
The first client to say yes is the one with the shortest decision process, which is frequently the smallest. They receive four enquiries a month, the assistant answers three of them, and nothing measurable changes.
That becomes the reference case. The agency now has evidence that it does not do much, and the clients who would have benefited most never hear a convincing argument.
Filter 1: enquiry volume
The first and simplest test. A client fielding a handful of enquiries a month will see nothing worth measuring regardless of how well it works.
Filter 2: repetition
Whether the same questions arrive constantly, which is what makes automation useful rather than merely present.
Filter 3: the cost of a slow reply
Whether losing an enquiry costs something identifiable, which is what turns a nice improvement into a business case.
Filter 4: when enquiries arrive
A client receiving substantial evening and weekend traffic has a gap that nothing else fills, which is the clearest argument available.
Filter 5: whether the answers exist
A client who cannot state their own prices, process or coverage is a client whose assistant will be vague, and the project will be blamed for it.
Filter 6: whether they will maintain it
Or pay the agency to, since an unmaintained assistant is confidently wrong within a season and becomes a complaint.
Filter 7: whether a person is genuinely required first
Some sales need a human from the first word, and proposing automation into those is how an agency loses credibility.
Filter 8: whether the client has something at stake
Clients under pressure to grow engage properly. Clients who are comfortable treat it as an experiment and do not maintain it.
Filter 9: whether the relationship is good
The first two clients should be ones who will be patient during the first fortnight and honest about what is not working.
Choose first: the client with volume and a clear loss
Which produces a reference case with numbers, and makes every subsequent conversation considerably easier.
Choose second: a different sector
So the agency has two cases rather than one, and can speak credibly to more of its base.
Decline: politely and with a reason
A client told honestly that this would not do much for them trusts the agency more, and is the judgement in agencies selling chatbots as a service to local clients.
Filter 10: whether anybody there will read the reports
A client with nobody responsible for marketing will not engage with a monthly report, which means the service becomes invisible and is cancelled at the first budget review regardless of how well it worked.
Choose third: a client who will talk about it
A reference case is only useful if somebody is willing to say so publicly. Establishing that willingness before starting costs one question and is frequently forgotten entirely.
Qualify the client before proposing anything
An agency that starts with the client who has volume, repetition and an evening gap gets a reference case with real numbers, where one that starts with whoever said yes first gets a case that proves nothing. The commercial model sits in pricing assistant builds for clients, the delivery in reselling assistants to your own clients, and the sector case in how marketing agencies use chatbots for their clients. See it working on the demo, compare on pricing, and the sector view sits under agencies.