I run scenarios for a handful of agency clients, and something keeps coming up that I don’t have a good answer to.
Client notices something in their CRM or inbox, asks me what the automation did and whether it was supposed to do that. I end up walking them through scenario history, which means nothing to them, or screenshotting it into an email. Takes ages and still looks unconvincing.
It’s got worse as I’ve added AI steps, because now the thing makes decisions I can’t always predict either.
So: does this happen to you? Or do your clients mostly not care as long as things work?
And if it does happen — is anyone sending clients something better than raw logs? The part I find hardest is when a client’s setup spans Make plus a couple of other tools, so there’s no one place to point at.
Possibly I’ve just got unusually nosy clients. Would be useful to know.
Raw scenario history is the wrong thing to hand a client. They don’t care that module 4 ran, they care what happened to their lead and whether it was supposed to. And it gets harder the moment there’s an AI step in the loop, because now you’re also explaining a decision you didn’t make by hand.
So rather than write about how I’d approach it, I built the thing I’d want to send a client. It’s running here on invented sample data, with no login: https://automation-digest-demo.9592.tech
It’s a weekly digest for one made-up client. Each automation gets a plain line about what it did for them that week. The part aimed at your hardest case is the cross-tool view: Make, HubSpot, Gmail, and Stripe all show up in one timeline, so there’s a single place to point at instead of four dashboards. And there’s a section that lays out every AI decision, the enquiry that came in, what the AI decided, why, and how confident it was, including one it wasn’t sure enough about and correctly left for a person.
It reads from the execution data your scenarios already produce, plus the recent events in the connected tools, so nothing changes inside your scenarios. There’s a short “under the hood” part at the bottom that shows the real run records it works from, if you want to see how it’s wired.
The page and the sample data are yours to look at either way. If it’s close to what you’re after, I’d be glad to set it up against your own clients’ scenarios.
@Nazar
hey! Christo’s digest demo is super clean, but if you want something lightweight that you can implement inside Make in 10 minutes without hosting a custom web app, here’s how I handle this for my agency clients:
I create a centralized “Audit Log” scenario. At the end of every client workflow, I add a webhook module that sends a simple payload (client_id, workflow_name, status, summary). If there’s a fuzzy AI step, I have the LLM output a 1-sentence explanation of its decision (e.g., “Assigned lead John Doe to Sales Tier 2 because budget > $5k”).
All of these hit a single Airtable base where I have client-specific views shared with each agency. They get a real-time, plain English feed of what ran, why it ran, and the outcome, with zero technical noise.
Drop me a DM if you want a quick screenshot of how I wire the audit log template!
Gabriele’s audit-log table is the right base — I’d push it one step further: clients don’t actually want a log, even a tidy one. Been through this with a trades-company client, same pain. What they want is two things only: what went wrong (and who handled it), and what the automation quietly saved them. So every branch of every scenario ends in one plain-language line written to that table — AI steps included, they append a one-sentence “why” for their decision. Then a tiny weekly scenario turns the table into a three-number email with an exceptions list. Nobody has asked me for screenshots since. The hard part: decide this at build time. Retrofitting sentences into a live scenario is the painful bit.