01 the business
Freight software provider
45 staff, 15 new customers a month
After a contract is signed, onboarding runs from one person's spreadsheet. Customers wait for logins and forget to send their data files. The first sign of trouble is often an unpaid first invoice.
02 before
The problem
After the contract is signed, onboarding runs on one person's memory and a checklist in a spreadsheet. Customers wait for logins, forget to send files and go quiet. The first sign of trouble is often a cancellation or an unpaid first invoice.
- Kickoffs are booked days after signing.
- Missing data files hold up go-live for weeks.
- Stalled accounts are noticed too late.
03 the build
What the AI does
- Step 1: Starts an onboarding plan when a deal is marked won in HubSpot or Salesforce.
- Step 2: Books the kickoff call and sends a welcome email listing what the customer needs to provide.
- Step 3: Tracks returned documents, data files and setup tasks, and sends polite reminders for what is missing.
- Step 4: Answers common setup questions in Intercom from your help centre and past onboarding notes.
- Step 5: Flags accounts that stall, such as no login after a week, to the account owner in Slack.
paused: waiting for a person to approve
How it was put together
- Started an onboarding plan when the deal was marked won.
- Booked the kickoff and sent a welcome email listing what the customer must provide.
- Tracked documents and setup tasks and sent polite reminders.
- Flagged accounts with no login after a week to the account owner.
04 human in the loop
Where a person stays in charge
The account owner reviews the plan for each new customer and decides when to step in on a stalled account. Any change to scope or price stays with them.
- The account owner reviews each plan and decides when to step in.
- Scope and price changes stay with people.
- Reminders stop when the customer replies.
05 rollout
From kickoff to live in 4 weeks
- Week 1Turn the onboarding spreadsheet into a standard plan per product.
- Weeks 2–3Run for new customers with the onboarding lead watching every step.
- Week 4Switch on stall alerts and the setup question assistant.
06 after
What changes
Every new customer moves through the same clear steps, and nobody falls through the gaps between sales and delivery. Problems show up in week one, not month three.
Assume 15 new customers a month and 6 hours of coordination each, falling to 2 hours with a tracked plan.
15 × 6 hours = 90 hours a month today. 15 × 2 hours = 30 hours. 60 hours a month go back to the onboarding team.
What to measure
- Days from signature to go-live.
- Customers active in their first two weeks.
- Onboarding tasks overdue.
- First-invoice payment rate.
07 evidence
What the research says
About 75% of generative AI's potential value sits in four areas: customer operations, marketing and sales, software engineering, and R&D.
Across 5,179 customer support agents, an AI assistant raised issues resolved per hour by 14% on average and by 34% for novice and lower-skilled agents.
NBER, Generative AI at Work (Brynjolfsson, Li and Raymond) (opens in a new tab) · 2023
08 lessons
Mistakes to avoid
- Too many emails. Customers need a clear list, not a drip campaign.
- No owner. Automation tracks; a person still owns the relationship.
- Measuring tasks, not outcomes. Go-live and first use are what matter.
09 questions
Common questions
- Does it work for professional services?
- Yes. Accounting and law firms use it for engagement letters, ID checks and document collection.
- Which tools does it use?
- Your CRM, a scheduling link, your help centre and a task tool such as Asana or Karbon.