A prospective customer asked for a quote in April. You sent it, answered a question and then got busy. The enquiry is still in the CRM, marked open. Somewhere in the mailbox there may be a reply, a postponement or a clear decision to go elsewhere.
Before spending more to attract another enquiry, it is worth finding out what happened to the ones already received. Your business has already invested in getting those conversations started, through advertising, referrals, events or somebody’s time.
AI can help reconstruct the history and prepare a useful follow-up list. The value comes from finding a credible next step, supported by the records. An old contact on its own is a weak reason to send another message.
Why this is a timely use of AI
Connected business tools give agents more useful context than a pasted contact list. Anthropic’s Claude for Small Business announcement includes HubSpot alongside document and workplace tools, with lead triage among the described workflows.
That is a product capability announcement, not evidence of recovered revenue. Your result will depend on what your systems contain, which connections are available and how carefully somebody reviews the proposed actions. You can also implement this approach with other tools or a controlled export of the relevant records.
The practical change is that reviewing several sources together can become part of a repeatable process, rather than a research task the owner keeps postponing.
Start with a small, defensible selection
Choose one group of records, such as quotes sent during a particular quarter that remain open. Include the business reason for selecting them and exclude records your team already knows should not receive further sales contact.
Keep the initial scope narrow enough to check every result. A pilot of 20 to 30 enquiries is an example of a manageable starting point, not a statistical guarantee or a universal recommendation.
Before analysing anything, confirm that the people and systems involved may access those records. Limit the assistant to the selected business sources. It does not need unrelated employee correspondence or a complete mailbox history to review one set of quotes.
Keep the originals intact. The first output should be a separate review list so mistaken interpretations do not silently change your sales records.
Reconstruct the conversation before scoring it
For each enquiry, collect the original request, the quoted scope, the latest relevant exchange and the recorded outcome. Match records using reliable identifiers where possible. Similar company names or a shared surname are not enough to join two customer histories.
The assistant should answer a few concrete questions. What did the customer ask for? What did your business promise? Who was supposed to act next? Is there a newer reply in a different thread? Was the quote superseded, accepted, rejected or left unresolved?
Require links to the underlying evidence. A missing reply should be described as absent from the sources checked, rather than proof that the customer never responded.
If the records conflict, put the enquiry in a clarification group. An honest unresolved case is more useful than a confident ranking built on an incomplete history.
Give the reviewer a reason to act
A useful shortlist has three outcomes: review for follow-up, resolve the internal record, or take no action. Avoid pushing every enquiry toward another sales message.
Consider three illustrative cases:
- A customer asked to revisit the proposal after moving premises. The stated month has arrived. Check the current circumstances before drafting a relevant follow-up.
- A quote appears open, but a later email confirms that another supplier won. Correcting the internal status may be the only useful action.
- A customer asked a technical question that your team never answered. Assign that question to the right colleague before suggesting a new pitch.
Each case should include an owner, the evidence, a proposed next step and any uncertainty. If the assistant assigns priority, require it to explain the reason in ordinary language. An unexplained score of 87 does little to help someone decide.
Draft from the actual history
Once a person accepts the proposed next step, the assistant can prepare a short draft that reflects the customer’s request. It should use current scope and availability, avoid invented familiarity and leave unsupported details out.
An expired quote needs a pricing review before anyone promises to honour it. A request involving a complaint may need a service response before any sales discussion. A previous purchase does not automatically make every new marketing message appropriate.
Check the applicable contact permissions and marketing requirements for the intended channel and country with the responsible person. Keep suppression records effective across the workflow. This article proposes an internal review process; it does not establish that a particular contact may lawfully receive outreach.
Immediately before any approved message is sent, check whether the customer has replied or a colleague has already contacted them. A list prepared on Monday can be out of date by Thursday.
Measure what the review produces
Count records reviewed, useful next steps, qualified conversations and work actually won. Keep corrections to the CRM separate from new sales opportunities. Both may be valuable, but they answer different business questions.
Track the staff time spent gathering records, reviewing suggestions, correcting errors and handling responses. Include the tool and implementation costs. A large list that consumes more attention than it saves is a poor result, even if the writing looks polished.
If a later order follows a reviewed enquiry, record the connection without claiming that AI caused the sale. For a larger evaluation, compare with a similar group handled through your normal process, while keeping customer treatment appropriate.
Revenue is also an incomplete measure. An illustrative €3,000 order with €2,400 in direct delivery costs contributes €600 before other costs. That figure is more useful for evaluating the effort than celebrating the entire €3,000 as a gain.
Make the process easy to stop and repeat
A named salesperson or owner should accept responsibility for the list. Store the review date and the sources checked, record what happened to each suggestion and prevent the same closed enquiry from reappearing every week.
Keep the first cycle manual at the sending stage. Once you understand the useful cases and failure modes, decide whether a recurring internal review earns its place. The aim is a smaller, better supported action list that fits the team’s capacity to respond.
XYZ by FORMATION can help connect the relevant sources, define the selection rules and implement review and duplicate checks. Our Sales Follow-up Operator is a starting point for discussing that workflow.
Bring a small set of unfinished enquiries and the question you want answered. We can help turn scattered history into a practical review process with clear ownership and measurable results.
Product description checked on 12 October 2026. All customer scenarios, pilot sizes and financial figures are illustrative. No customer revenue result is claimed.