What to automate in customer onboarding, and what needs a person

Every company that scales past a handful of people eventually asks the same question: what should live in a system, and what needs a person to carry it. Get this wrong and you either drown your team in meetings that could have been an email, or you automate away the judgement calls that actually protect the customer relationship.

The panel on Session 3: Mastering the Sales-to-CS Baton Pass worked through this directly, using an air traffic control comparison that gives a useful, testable rule rather than a vague principle.

The short version: structured, stable facts belong in a system. Anything involving judgement, nuance or a relationship needs a human. But that rule only works if the data underneath it is actually usable, which is where most organisations quietly fall down.

What's the test for automating a handoff?

Angela Sutton described the split as being like air traffic control: radar tracks altitude and speed continuously, and that data can run on its own. But when two planes are converging and about to land, a human voice takes over. The system feeds the moment; it does not replace the decision.

Erin Felton and Anderson Campbell both landed on a similar version of this rule. Felton put it as systems carrying the structured, repeatable, reportable information, like revenue size or firmographics, things that will still be true in six or nine months. Humans carry the nuance: decisions, relationships, anything more fluid. Campbell added that this holds regardless of whether you're looking at internal handoffs, external handoffs, or customer marketing. The line isn't about scale, it's about whether judgement is required.

Campbell also made the point that automated systems can be useful for triggering human attention rather than replacing it. The facts from a system can act like an alarm bell that tells a person where to look, without making the call itself.

Why does data hygiene have to come first?

Kristine Kukich was blunt about this: bad data in, bad decisions out. None of the systems-versus-people logic works if the underlying data about a customer hasn't been aggregated and cleaned. She noted that this is one of the few genuinely good side effects of the AI push in customer teams: it has forced a lot of organisations to finally run data hygiene projects they'd been avoiding, because AI tools expose bad data faster than a person quietly working around it ever did.

This connects to a point Angela Sutton raised earlier in the same discussion: research shows that asking customers to repeat themselves is a top driver of churn during onboarding, alongside slow onboarding and poor communication. A system built on shaky data is one of the more common ways customers end up repeating themselves, because the facts that should have travelled with the account never made it into a usable form.

Where does a dedicated context role fit in?

Jamie McDermott raised the idea of a dedicated information and context manager, someone who sits between sales, CS, onboarding and support, keeps the living documents current, and knows what each team actually needs to know. The panel liked the idea in principle. Erin Felton noted she's seen similar roles exist between marketing and sales, but that they tend to eventually get absorbed into whichever department they end up reporting into, losing the neutrality that made them useful in the first place.

Kristine Kukich argued the role works better sitting inside the organisation, under something like the CCO or CX function, with genuine AI fluency, rather than as an outside consultancy, so the institutional knowledge stays put. McDermott suggested a tightly mandated third-party consultancy could preserve that neutrality instead. Neither version was resolved on the call, but the panel agreed on the underlying need: someone has to own the flow of context, or it defaults to whoever remembers to write things down.

How should the internal handoff meeting actually work?

Alison Bukowski's point about the handoff itself is really a version of the same systems-versus-people logic. She argued that most organisations try to save time by skipping the internal conversation before a handoff and leaving notes instead. Her fix was two meetings: an internal one where sales shares context and color commentary with the post-sales team, and an external one where sales stays in the room so the customer feels supported.

1. 1
   Capture facts before the call

   Structured details like buying motive and success criteria go into the system so no one re-asks them.
2. 2
   Hold the internal meeting

   Sales shares context, color commentary and anything that doesn't fit neatly into a field.
3. 3
   Keep sales present externally

   The customer meets the new team with the old team still in the room, not a cold handoff.

Erin Felton added a detail worth keeping: in that external meeting, it has to be obvious to the customer who everyone is and why they're there, not just a string of names and titles. That's a judgement call about communication, not a data problem, and no system fixes it.

Frequently asked questions

What should never be automated in a customer handoff?

Anything involving judgement, relationship nuance, or a deviation from the standard path, according to the panel. Facts can move through a system; decisions about what those facts mean for a specific customer need a person.

Why does AI adoption often surface data problems?

Kristine Kukich pointed out that AI tools expose bad underlying data much faster than manual workarounds did, which is why many organisations only started serious data hygiene projects once they tried to put AI to use.

Should companies create a dedicated context or handoff role?

The panel was positive about the idea of a person who owns context flow between teams, but flagged a real risk: such roles tend to get absorbed into whichever department they report into over time, unless there's a deliberate structure to keep them neutral.