Amy
Axelrod

  • Sum of All Parts
  • All There
  • The Garden of Forking Pawns

Sum of All Parts: Simplified Customer Onboarding

This case study examines a common problem: a simple customer UI can mask a network of operators with different responsibilities and the processes those entail. When automation reduces the process burdens, judgment and attention can both be leveraged to create lasting value.

Please email amydaxelrod [at] gmail [dot] com
for resume requests.

LinkedIn
Click to enlarge
1 / 5·

Notes

Independent concept based on experience in customer onboarding. Illustrative data. No confidential information.

Reasoning

The problem isn’t the people, and it isn’t a missing source of truth per se. The problem is that each person is their own source of truth, siloed in their own role. To accomplish one thing, everyone has to coordinate multiple steps instead of the work moving seamlessly. It’s a system limitation, not a lack of effort.

So the answer isn’t more hands. It’s a better system, one where information moves more freely and the partner doesn’t have to carry that cognitive burden.

The thing worth protecting is the customer’s experience. The rest can run in the background, so the partner can focus on judgment.

References

Donella Meadows (1941–2001) was an environmental scientist and pioneer of systems thinking. Her 1999 essay “Leverage Points: Places to Intervene in a System” remains a cornerstone of the field. She expanded it in Thinking in Systems: A Primer (2008), published after her death, and it is still taught in university courses on systems design today.

In the essay, she defines leverage points as places within a complex system where “a small shift in one thing can produce big changes in everything.” She ranks twelve, from least to most effective. The sixth, the structure of information flows, is the most relevant here: who has access to what information, and when. Meadows observes that “missing feedback is one of the most common causes of system malfunction,” and that restoring it often achieves more than adding people or effort, which rank among the weakest levers. In customer onboarding, feedback can be fragmented or manual, especially among teams with different functions. The solution that follows addresses these information flows, smoothing them so the partner no longer has to search for or relay updates between teams, when the role is best exercised with the customer.

The design also touches two higher points: its rules (what the system may send on its own) and, on the next slide, its goal (the customer should need only one person).

Donella Meadows, “Leverage Points: Places to Intervene in a System” (Hartland, VT: The Sustainability Institute, 1999), https://donellameadows.org/archives/leverage-points-places-to-intervene-in-a-system/.

Donella H. Meadows, Thinking in Systems: A Primer, ed. Diana Wright (White River Junction, VT: Chelsea Green Publishing, 2008).

Keywords: process mapping · root-cause analysis · systems thinking

Independent concept based on experience in customer onboarding. Illustrative data. No confidential information.

Reasoning

When you strip the role down, three things remain: the customer's experience, preparing for activation, and keeping the process excellent. Everything else is coordination: status questions, duplicate updates, nudges, scattered docs. It's necessary, but it isn't the job.

The customer shouldn't see the workshop that exists behind their experience. "Let me check with another team" isn't a people problem, it's a design problem.

This isn't a trade-off. Every team functions, and the customer experience improves as a result.

References

First-principles thinking means breaking a problem down to what has to be true, then building back up from there instead of copying how it's always been done. Applied here, it separates the work only the partner can do from the work a system can carry. Farnam Street's overview is a clear introduction to the method.

Shane Parrish, "First Principles: The Building Blocks of True Knowledge," Farnam Street, https://fs.blog/first-principles/.

Keywords: first-principles thinking · customer experience · process excellence

Independent concept based on experience in customer onboarding. Illustrative data. No confidential information.

Reasoning

The path from contract to activation isn't a straight line. It's more like a tree, and every change branches out to a different team. One change shouldn't have to be carried to each of them by hand.

So reviewing updates, drafting, and tracking aren't three tools. They're one layer. The layer handles what it can and hands back anything that needs judgment. The record, the customer's inbox and final approval stay with the partner. Only acknowledgements go out on their own, and each one is logged.

Routing alone would just move the noise somewhere else. The value is in context: knowing what a message means and who needs to see it. Three scheduled digests a day replace checking channels all day, and a hardware change that used to touch several systems now takes one checkbox and one approval.

Risks and Mitigations

RiskMitigation already in the design
Outdated or wrong status updateStatus is set by the partner; consistency check
Over-trusting draftsHuman approval, change log, thumbs up/down
Approval fatigueAuto-acknowledge only, batched queue
Access and privacyScoped channels and docs, no record-system access
Out-of-office coverageScale path → shared context

Considered and Rejected

  1. Connecting the AI directly to the record system. It's the most complex build, needs the most compute, and carries the most permissions risk. The partner keeps records accurate by hand, and the record system's limits are their own problem.
  2. Auto-sending every reply. One wrong status costs more trust than the time it saves. Only acknowledgements are sent automatically, and each one is logged.
  3. Letting the AI send customer emails. The customer relationship is the human work. The AI drafts from approved scripts; the partner sends.
  4. Letting the AI set customer status. Status is a judgment call. The partner sets it; the AI reads it and acts on it.
  5. A new standalone app. One more place to check works against "one person, one place," and adoption is slow. Built from tools teams already use.
  6. Real-time alerts for every message. They bring back the noise the system is meant to remove. Three digests a day instead, with urgent items flagged, and exceptions considered for extremely urgent updates.
  7. A rules-only automation. It can route messages but can't draft in context or judge what needs the partner, and that's where the time goes.
  8. Adding a coordinator. It moves the relay to another person instead of removing it, and it adds cost.

References

The reveal, three things that turn out to be one, is modeled on Steve Jobs's 2007 iPhone launch, where he introduced an iPod, a phone and an internet communicator before revealing "these are not three separate devices, this is one device."

Lisa Eadicicco, "Steve Jobs's iPhone Launch Keynote, 2007," Time, https://time.com/4628515/steve-jobs-iphone-launch-keynote-2007.

The swimlane layout borrows from the service blueprint, a design method that maps what the customer sees (front stage) against the work that makes it happen (backstage).

Sarah Gibbons, "Service Blueprints: Definition," Nielsen Norman Group, https://www.nngroup.com/articles/service-blueprints-definition/.

Keywords: internal tools · workflow automation · AI workflows · grounded context · prompt library · Apps Script · human-in-the-loop

Independent concept based on experience in customer onboarding. Illustrative data. No confidential information.

Reasoning

The point isn't saving time. It's getting attention back. Without the relay, the partner can become an expert and an advocate, with their attention on the customer instead of the inbox.

The customer gets one point of contact and a real chance to learn the platform before activation. Teams get acknowledged right away, with complete requests and a clear view of what's waiting.

The organization gets a process that's standard enough to improve. With this many moving parts, it's hard to know what to change. Simplify one thing, and real improvement becomes possible. All of this still has to be proven in a pilot.

Measurement Plan

  • Log one baseline week.
  • Run 5–10 pilot partners for the length of a typical activation cycle, compared with the rest of the team.

Measure

  • Manual updates per change
  • Response time to teams
  • Scheduled touchpoints on time
  • Draft approval rate
  • Whether partners keep using it after the pilot

Limits

  • It's a one-person tool until it's shared.
  • Review time is the ceiling.
  • Status is only as fresh as the last update.
  • Access decisions are needed.
  • The process-doc backlog is its own project.

Scale path: one person → pilot team → shared context. All figures are illustrative until the pilot.

References

The process follows the double diamond, the Design Council's model for design work: discover and define the right problem, then develop and deliver the right solution. Each diamond opens wide before narrowing to a decision, which is why the deck spends as long on the problem as on the fix.

Design Council, "The Double Diamond," https://www.designcouncil.org.uk/our-resources/the-double-diamond/.

Keywords: time saved · error rate · adoption · change management · pilot · metrics

Independent concept based on experience in customer onboarding. Illustrative data. No confidential information.

Reasoning

I didn't want to organize this by tool, because that only tells you what was used. Organizing it by method shows how the decisions were made.

The judgment stayed with me: what to look at, what mattered, what to keep. AI widened what I could see and sped up the work, but the choices were mine.

References

W. Edwards Deming (1900–1993) was the statistician whose work shaped modern quality management, from postwar Japanese manufacturing to the American quality movement. He held that most problems come from the system, not the individual worker, an idea often paraphrased as "a bad system will beat a good person every time." That idea runs through the Discover row: a system problem, not a people problem.

"W. Edwards Deming," Wikiquote, https://en.wikiquote.org/wiki/W._Edwards_Deming. (Paraphrase; no confirmed original source for the exact wording.)

Keywords: AI fluency · responsible AI use · human-in-the-loop

·

Notes

Click the slide to zoom