Key Takeaways
The email goes out at 09:00 on a Thursday: “We noticed you haven’t logged in for 14 days. Here are three features to try.” It reaches the operations lead at an industrial-equipment customer. On Wednesday she spent 40 minutes failing to find the firmware article for the model her team installed last month, gave up, and opened a ticket. The journey fired on schedule. The knowledge she needed wasn’t there. Two systems, both called digital customer success, and only one of them knew she existed.
That email is the artifact most digital CS programs are built around: the automated touch. It’s cheaper than a CSM’s call and it scales. It also does nothing the customer asked for. The distinction between automating the touch and removing the need for one is the whole subject, and the data on where programs stall makes it hard to avoid.
Digital customer success is the least-built strategy in CS, and the numbers say why
The adoption story everyone quotes comes from Gainsight’s CS Index: digital tools such as communities and self-service portals jumped from 42% of companies in 2023 to 73% in 2024. Treat that with care. A 31-point swing in a year, in a vendor survey of about 250 companies, is far more likely a change in who answered than a change in the market. The number that describes what’s actually built is ChurnZero’s. Asked which CS strategies they had put in place, 1,027 leaders put digital customer success at 17% “fully implemented”, 46% partial, 19% planned and 19% not at all. Tied for last with voice-of-customer. Compare the 54% who had fully separated technical support from account ownership.
Then look at who is supposed to build it. Dedicated CS Ops roles, the people who design journeys and wire data, fell from 61% of teams in 2023 to 54% in 2024, and 41% in ChurnZero’s 2025 study. Budgets went the same way: the share of leaders reporting a budget increase fell from 52% in 2022 to 37% in 2023 to 25% in 2024. In 2023, 55% of CS teams held or cut headcount, and 15% of companies over 1,000 employees cut CS staff, up from 4%. The Customer Success Collective reports CS job-opening growth at an all-time low and 71% of CS professionals job-searching.
Put those together and the pattern is plain. Digital customer success is being bought as a substitute for the people who were just let go, by teams that also let go of the people who would build it. That predicts exactly the “partially implemented” plateau the data shows: a journey builder licensed, three sequences live, nobody left to instrument the rest. And it explains the ratio data. 62% of CSMs carry 60 accounts or fewer; 3% run books of 500-plus. Genuine tech-touch scale is rare. Most “digital” programs run alongside an unchanged book, sending the Thursday email to accounts a CSM already owns.
Two different things are called digital customer success
Strip away the vendor language and the programs split cleanly.
Outreach automation puts the CSM’s calendar on rails: lifecycle emails, usage nudges, renewal reminders, in-app announcements, NPS campaigns, health-score alerts. Gainsight’s 2026 priority list reads as exactly this: advocacy and NPS at 54%, feature adoption at 44%, training at 42%, onboarding at 40%. It cuts the cost of a touch from a call to a send. It doesn’t change what happens after the customer opens it.
Customer enablement changes what happens after. The customer finds the firmware article, completes the setup, answers her own question, and never generates the contact at all. Nobody reached her because she didn’t need reaching.
The difference is structural, not stylistic. Automating a touch makes a linear cost cheaper: every new account still gets its sequence, and the sequence still lands on a person who has to do something with it. Enabling a customer removes the touch permanently, for that account and for the next thousand with the same question. Automation makes scaling cheaper. Enablement is what makes it compound, which is why scaling CS without scaling headcount is an enablement story rather than a journey-builder story. The test for any item on the digital CS roadmap: does it schedule the next contact, or remove it?
The evidence for enablement is stronger than the evidence for outreach
There is one randomised experiment in this entire literature, and it’s about enablement. Retana, Forman and Wu worked with a public cloud provider on 2,673 new customers, 366 of whom received proactive guidance on basic features in their first days. The treated group churned at half the rate in the first week, asked 19.55% fewer support questions in week one, and used 46.57% more of the service over eight months. The support effect faded within a week. The usage and retention effects didn’t. That’s the shape enablement has: a one-time act that keeps paying.
It also tells you where the lever sits. Not in the renewal quarter. In the first week, on the basic task, before the customer has formed a habit or a grievance. That’s onboarding at scale, and it’s the part of digital CS most programs fill with a checklist email.
The outreach side has no equivalent study, and the self-service data explains why nudges underperform. Gartner surveyed 5,728 customers: 14% of issues were fully resolved in self-service, 36% of the ones customers called “very simple”. The reasons weren’t engagement. 43% couldn’t find content relevant to the issue; 45% felt the company didn’t understand what they were trying to do. Customers want to self-serve: ServiceNow’s survey of 27,000 consumers found 75% try self-service before calling, and 87% still name the phone as their preferred channel. The first attempt is digital. The resolution isn’t, because the content underneath the surface is missing or stale.
One more number for the feature-adoption campaign in the priority list. Pendo’s benchmarks across 6,800 products found 6.4% of features drive 80% of clicks. A nudge campaign for the other 93.6% is a campaign to promote features the market has already declined. The binding constraint on digital CS isn’t the journey engine. It’s whether the answer to the customer’s actual question exists, is current, and is where she’s looking. That’s why self-service plateaus at roughly the level Gartner measured.
How the industry counts digital customer success, and why the number can’t tell you
Before setting a target, read the definition behind the metric your vendor reports. Intercom’s documentation for its Fin agent defines two kinds of resolution: a “confirmed resolution”, where the customer says it helped, and an “assumed resolution”, where the customer exits the conversation without requesting further assistance. Assumed resolutions are billable. A customer who gives up and closes the tab is counted, and charged, as a solved problem. That one definitional choice means the headline automation metric in the category can’t distinguish a resolved customer from a defeated one.
Klarna is the case study both sides quote, and both halves are real. In February 2024 its assistant handled 2.3 million conversations in its first month, two-thirds of all chats, with a 25% drop in repeat inquiries and a projected $40 million profit improvement. By May 2025 the company was bringing human agents back in a “dual-track” model, which it described as a small pilot rather than a reversal. The lesson isn’t that AI failed. It’s that the metric that made the first announcement possible, volume handled, said nothing about whether the customer was done.
So measure digital CS on two numbers the vendor dashboard doesn’t default to. Confirmed resolution, where the customer or the next 30 days of silence says the issue ended. And repeat-contact rate by account. The Thursday email’s recipient is a repeat contact waiting to happen. A program that watches those two numbers finds out in a month whether it’s enabling or just sending, and the metrics that change a decision stay a short list.
What to automate, what to enable, what to keep human
Run every recurring CS activity through three questions, in order. Does it remove the next contact, or schedule it? Is the output the customer’s outcome, or your team’s calendar? What does a wrong call cost, and can it be undone?
| Activity | Removes or schedules the contact? | Verdict | What it needs underneath |
|---|
| First-week guided setup for a new account | Removes | Enable, then automate the delivery | Current setup content per product and model |
| Answers to the 50 questions every account asks | Removes | Enable: help centre and in-app assistant | Structured knowledge the assistant can read |
| “Haven’t logged in for 14 days” nudge | Schedules | Automate only if it links to the answer the account was searching for | Search and ticket history per account |
| Health-score alerts | Schedules | Automate the alert; a person reads it | Usage, requests and onboarding status on one record |
| Renewal reminders on flat usage | Schedules | Automate | Contract and usage data |
| Feature-adoption campaign | Schedules | Skip unless the feature is one the account’s users asked about | Search-gap data by account |
| Multi-stakeholder renewal, champion change, escalation | Neither | Keep human | Full account history in front of the person |
Two things the table makes visible. First, the enablement rows sit at the top because they decide how many of the lower rows you need. An account that found its answer in week one doesn’t receive the 14-day nudge. Second, every row depends on records: which model the account installed, what its users searched for, where onboarding stalled. When those live in four tools, the automated tier runs on a stale export, and the first bad nudge to a good customer is the day the CSMs stop trusting the program.
On the human row, be honest about what the data supports. ChurnZero found companies that meet customers in person have better NRR, then added the caveat most people drop: “companies that sell larger deals invest more in relationships”, and the effects were moderate to small. Touch model doesn’t cause retention. Contract size decides which touch model pays for itself. The right question is not whether digital retains as well as human, but at what ACV a named CSM stops earning back her cost, and no published study answers it. Set that line from your own numbers.
A 90-day digital customer success plan that starts with the knowledge, not the journey
Most programs start by licensing the journey tool and writing sequences. Start at the other end.
- Weeks 1–3: find the 50 questions. Pull a quarter of tickets, help-centre searches and assistant transcripts. Cluster them. The top 50 are the enablement backlog, and most have no current answer.
- Weeks 3–6: write the answers once, structured. Product, model, audience, region as fields, not folder names, so the same article serves the help centre, the in-app assistant and the agent’s reply. Ship the first-week setup guide first; it’s the one with a randomised trial behind it.
- Weeks 6–9: wire the journeys to the knowledge. The 14-day nudge now links to the article the account’s users searched for and didn’t find. The health alert shows the search gap. The onboarding sequence knows which milestone stalled.
- Weeks 9–13: measure confirmed resolution and repeat contacts by account. Not sends, opens or assumed resolutions. Expand to the next tier only when both numbers move.
Step three is where most stacks break, because the journey tool, the help centre, the assistant and the ticketing system don’t share a record. In MatrixFlows they do: the article, the account, its requests and its onboarding milestones are one linked foundation, and the help centre, the in-app assistant and the Conversations Inbox read from it. The operations lead’s failed Wednesday search shows up as a gap against her account and against the firmware article that doesn’t exist yet. Someone writes it once. Thursday’s email, if it still sends, links to it. The next account with that model never searches at all. That’s the loop that compounds, and the journey engine is a small part of it.
Digital customer success isn’t automated outreach. It’s a customer who didn’t need to be reached.
Create a Free Workspace → and build the 50-question backlog before the next journey sequence.
.svg)
.svg)
.svg)
