The pitch is always the same. An AI answers your enquiries in seconds, writes the follow-ups nobody got round to, and tells you which deals are going cold. It is a good pitch, and the technology behind it is real.
Then you connect it to your business and discover what it has to work with.
Where the enquiries actually live
In most companies we meet, it is roughly this. Web enquiries land in a shared inbox that two people half-watch. Half the real conversation happens on WhatsApp, on the phone of whoever picked it up. Quotes are PDFs in a folder named after the year. The pipeline is a spreadsheet somebody updates on Thursdays, mostly.
No model can do anything useful with that, because the record of the relationship does not exist in one place — it exists in several, in fragments, with no connection between them. The thing people call "an AI problem" is almost always a data problem wearing a new hat. The proof is simple: a new salesperson joining on Monday could not answer "what happened with this customer?" either.
What it would need before it could help
- One contact record per person, with the whole timeline on it — the form they filled in, the texts, the booking, the invoice.
- Every lead tagged by source at the moment it arrives, not reconstructed later from memory.
- Conversations in one thread, not one per device and one per colleague.
- Pipeline stages with history, so "this deal is stuck" is a fact with a date on it rather than a feeling in a meeting.
- Tasks attached to the contact, so the next action survives the person who thought of it.
That is the unglamorous half of the job, and it is the half that decides whether anything clever works later. It is also what Hubb Stream is for: contacts with tags and a full activity timeline, visual pipelines where every stage change is recorded, and a unified inbox with two-way SMS over Twilio so the text thread is on the contact and not on a phone in someone’s pocket.