Once the CRM was in and sales follow-up was automated, another problem became visible. The company’s tools did not talk to each other. The accounting software knew about invoices, the CRM knew about deals, and somebody joined the two by hand every month.
This engagement removed that manual join, and took the opportunity to automate two or three other things nobody had time for.
Keeping accounting and the CRM in step
The first job moves information between the accounting software and the CRM: clients, quotes and invoices, each in the right direction, with the history loaded at the start.
The point is not just to avoid re-keying. It is that running the sales side stops being disconnected from reality. A deal marked won in the CRM has an invoice behind it in the accounts, and any gap between the two shows up straight away instead of surfacing at the year end.
Follow-ups that no longer depend on memory
The second job emails prospects according to the real state of the deal, rather than according to when somebody happens to think of it.
This is a standard piece of CRM automation, but it is worth more when the CRM is current. A follow-up built on wrong information irritates the person who gets it. A follow-up sent at the right moment, with the right context, gets read.
A welcome message written for each person
The third job is the most interesting, because it shows what generative artificial intelligence really adds to ordinary automation.
When someone subscribes to the newsletter, a model writes a personal engagement message from what is known about them: their sector, their company, the content that brought them in. Not a template with the first name dropped in, but a message about their subject.
The difference in engagement is clear. A subscriber who receives a message about their own line of work replies, and some of those replies turn into sales conversations. This is exactly the kind of task no team can do by hand at scale, and that a model does well.
The part everyone forgets: monitoring
All of these jobs are deployed on the cloud and run unattended. That is precisely why they are dangerous if nobody is watching them.
Automation that breaks quietly is worse than no automation at all. It gives the impression that the work is being done, and three months later you find out the sync stopped. So every job reports on its own execution, and anything abnormal raises an alert addressed to a named person.
It is the least impressive part of the engagement and it is the part that decides whether any of it is still running in two years.
What transfers
This architecture fits any pair of management tools with a programming interface, which covers every accounting package and every CRM on the market.
The mistake to avoid is trying to synchronise everything. Start with the flow that costs the most re-keying, make it reliable, monitor it, and only add the next one when the first has run for several weeks without anyone touching it.
The end client is not named. The figures quoted are those of the engagement.