Automatic lead qualification
Takes an enquiry, asks the right questions and qualifies intent, budget, urgency and location before passing it on.
The best automations do not start with “what can AI do?”, but with “what costs time or money here?”.
Takes an enquiry, asks the right questions and qualifies intent, budget, urgency and location before passing it on.
Collects the data, photos, location and size of the job, then prepares a structured request.
Chases prospects who stopped replying, within the applicable rules.
Handles enquiries received at night or at weekends and prepares the call-back for the team.
Answers questions, identifies the need and prepares the handover to a salesperson.
Avoids pointless meetings by gathering context beforehand.
Turns project data into a first proposal structure, for a person to review.
Handles repetitive requests and passes exceptions on.
Classifies urgency, department and subject.
Checks that the service solved the problem and surfaces blocking issues.
Collects the need and availability, then prepares the booking.
Confirmations, reminders and rescheduling.
Asks for feedback at the right moment and opens a path to a public review.
Spots negative feedback before any public review request and alerts the team.
Chases missing paperwork and tracks progress on the file.
Gathers all the information needed after the sale.
Processes emails, documents and forms, then organises the information for internal systems.
Collects the problem, location, urgency and availability, then prepares the intervention.
Pulls the data together and produces readable periodic reports.
Identifies status, delays and blockers, then notifies the people responsible.
These are opportunities to explore, not Agents available in the catalogue. We hold no data proving demand for any of them.
Both approaches are valid. They simply answer different questions.
Rather than a “quote request Agent”, build a “quote request Agent for roofers”. It can then understand precise vocabulary:
A general-purpose Agent can be superior when the problem is genuinely horizontal, for example:
A buyer almost never pays for “more features”. They often pay for something else: less work between installing the product and getting the expected result.
A highly specific Agent already knows the vocabulary, the workflow, the data needed, the exceptions, the questions to ask and the expected outcome. It therefore needs less customisation, which strengthens the sense of fit — even if the total market is smaller.
A particularly interesting approach keeps a common engine and presents it commercially by trade.
The core stays shared, while the prompts, the questions, the configuration, the onboarding, the demonstration and the copy are adapted to each trade.
This can combine the technical efficiency of a horizontal product with the commercial strength of a specialised solution. It is not always the right answer: it depends on the problem.
One solid Agent beats a library of automations that fail. Aim for a product that is:
Turn it into a workable automation, then into an offer.
I have an idea