Six tests to run before you buy client-memory software

Bring one real client history to the demo and try to load it — three charters, the wine the family drank, a child's birthday. Jennifer Kerum, co-founder of Anchor Croatia, a luxury yacht charter brokerage, says basic client-profile programs already exist "but not to the extent that I would like to have it." That single test eliminates most candidates in an afternoon.

What is client-memory software?

Client-memory software, sometimes called relationship intelligence or AI-native CRM, is a system that automatically builds and maintains a record of each client relationship — preferences, history, personal context and prior interactions — by reading the communications and documents a business already produces, then surfaces the relevant details before the next interaction.

A conventional CRM is a pipeline tool: it tracks deals through stages, and any relationship detail is something a person typed in. Client-memory software inverts that, and the unit of value becomes context rather than deal stage. For a business with a few large repeat clients rather than many small transactions, that inversion is the whole point.

What detail actually needs to be captured?

Experience detail, not transaction history. Asked which single AI tool would most change her job, Kerum did not name anything client-facing:

"A client comes on board and they spend a week with me. I really want to get really personal about what wine they're drinking, where they visited, their kids' birthdays. And I would love to have it where it's a really easy way to organize that information on the back end."

The commercial logic is retention — "the goal is to always keep your clients, right?" Artificial Reality host Galina Fendikevich named the customer-side value directly: if the next charter is stocked with the wine you drank last time, you stop shopping elsewhere.

Who is building this, and how far along are they?

A funded category of AI-native CRM now builds client records automatically from email, calendar and meeting transcripts, with no manual logging. Attio has raised $85M in total and is the best funded of that set. Day.ai, built by HubSpot's former Chief Product Officer, raised a $20M Series A from Sequoia. Mesh, formerly Clay, raised an $8M Series A from Sequoia and General Catalyst and is the one designed around personal context rather than commercial pipeline. Lightfield, from the founders of Tome and backed by Lightspeed, and folk, an Accel-backed lightweight CRM for service businesses, sit in the same space. Artificial Reality covers the wider memory layer in a separate guide.

All of them are horizontal. None models a charter, a vessel or a crew. Every one is closer to Kerum's request than a traditional CRM, and none is built for it.

Which six tests should you run before signing?

• Load a real client history. Take a family who has chartered three times and see whether the system can hold what actually matters about them.
• Ask where the memory comes from. If capture is limited to email and calendar, and your relationship detail arrives by WhatsApp and crew debriefs, the system will be confidently incomplete.
• Ask what surfaces, and when. A record nobody reads at the right moment produces no value. The requirement is that the wine appears when the next charter is being planned.
• Test the multi-user case. Several brokers touch one client. Confirm they all see and contribute to the same record.
• Ask about export before you sign. This dataset becomes your most valuable asset. If leaving means losing it, you have rented your own client relationships.
• Price the setup honestly. The more configurable the tool, the more it needs someone to configure it — and in a small team, that person is usually the owner.

Is this AI intruding on a human relationship?

The operator's position is the opposite. Kerum:

"A lot of people talk about how AI is taking away personal experience. But if you're able to use the tools to help build it even more, why wouldn't you?"

The system remembers so that a person can act on it. Nothing in the request replaces the broker.

What should you do this week?

Write the record before you buy the software. Take your ten most valuable repeat clients and put one page against each: wine, locations visited, family milestones, anything a crew mentioned. It costs an afternoon, it is portable, and it is the exact dataset you will use to test every product in this category — because the systems that cannot ingest it are the ones that would have failed you six months after signing.

Where this comes from

S1E2: How a luxury yacht company uses AI without losing the personal touch — the full interview with Jennifer Kerum. Listen or watch: YouTube, Spotify or Apple Podcasts.