How do we flag account risks and upsell opportunities?

Why risk and opportunity both arrive late

The client retention essay in the newer book describes surprise churn. Clients appear satisfied, delivery is solid, invoices are paid, and then the work stops or the contract is not renewed. It says the problem is not that founders failed to care but that the retention model they borrowed from software was built for a business that produces usage data, and services do not. In professional services the signals exist, it says, but they live in meeting conversations, email tone, responsiveness, executive presence and subtle shifts in engagement.

It explains why earlier fixes did not work. Relying on founders meant relying on memory. Adding account plans, reviews, surveys and client success roles added process without signal: health scores based on opinion, records based on self-reporting, reviews that were episodic and backward-looking, and surveys that captured a snapshot rather than a trajectory.

The account management essay describes the same problem on the opportunity side. Delivery staff only see openings when the client pushes them, so expansion becomes accidental rather than intentional, and white space goes unmapped.

Risk signals to watch

The retention essay lists the early signals AI can detect across meetings, emails, deliverables and cadence: disengagement, value confusion, relevance drift, executive withdrawal and priority loss. It adds the patterns people miss, such as subtle tone changes, declining participation, shifting stakeholder influence and reduced urgency or decisiveness. The account management essay adds relationship drift, leadership changes, scope creep, missed deadlines, competitor activity, vendor fatigue and champions who have become under-engaged.

Each revenue type fails differently

The retention essay says one of the most damaging assumptions is that retention is a single problem. It describes how commitment erodes in each model. Retainers rarely fail from dissatisfaction; they fail from perceived stagnation, when the client wonders whether the work is still evolving. Subscriptions churn from value opacity, when the client cannot say what they are getting. Outsourcing contracts erode through silent replacement, as another vendor or an internal team takes the work over. Fractional executive roles suffer executive drift as the leadership team matures or changes. Long-running projects suffer momentum decay as urgency dissolves.

As an inference, tag every account with its revenue type and watch for the failure that type tends to produce, rather than asking one generic question about whether the client is happy.

Structural risks a buyer would see

The client relationships chapter of the 2020 book lists what an acquirer checks. No single client should be worth more than 10 percent of billings. Average client tenure should be three years or more. Client quality matters, because revenue from start-ups is less reliable than revenue from large enterprises. The health of the client end markets matters too. And relationships must sit with the firm rather than with one key employee, documented in account plans in a system everyone uses.

As an inference, these belong on the same risk list. An account can be perfectly happy and still be a risk to the firm because it is too large, too new or held by one person.

Opportunity signals to watch

The account management essay describes AI finding new problems inside existing accounts, identifying projects the client has implicitly requested through questions or comments, spotting needs before the client articulates them, cross-referencing patterns across the whole client base, detecting moments when clients are most receptive to new ideas, mapping white space across all service lines, and prioritizing plays by revenue, timing and probability. It also flags when a competitor in the account is weak.

The service offering chapter of the 2020 book adds a warning: if you keep bringing clients the same thing, they become fatigued. As an inference, a client who has bought only one service from you for years is both an opportunity and a stagnation risk.

A leadership change is both

The 2020 book tells a story from SBI. Its clients were sales leaders, whose average tenure was about eighteen months, and when a sales leader left, the work typically stopped. The author worried a buyer would see short client tenure and walk away. His banker found the opposite in the data: departing sales leaders hired SBI again at their new companies.

As an inference, treat every leadership change as two flags. The account you have is at risk, so build the relationship with whoever replaces your contact. And the contact who is leaving is a potential new client, so stay close as they move.

Turn flags into action

As an inference from the material, a flag is only useful if someone acts on it. Keep one list per account holding both kinds of flag, each with a date, the evidence behind it, a named owner and a next step. Review the list for your most important accounts on a fixed rhythm, rather than when someone remembers. Match the response to the flag: the account management essay draws on four conversations credited to the book The Expansion Sale, which are why stay for renewals, why pay more for price increases, why evolve for new work and why forgive for recovering from a service problem. A value confusion flag usually needs a why stay conversation; a missed deadline needs a why forgive conversation; an implicit request needs a why evolve conversation.

The retention essay sets the stakes: until retention consistently exceeds roughly 90 percent, it says, scaling stays harder and more fragile, because new revenue replaces lost revenue rather than compounding on top of it.

Let the system watch and people decide

Both essays divide the work the same way: AI monitors continuously across every account and surfaces what matters, and people lead the conversations, re-anchor value, navigate executive change and decide. As an inference, the account manager should spend no time searching for flags and all of their time acting on them.

What we do not prescribe

Collective 54 publishes no health score, risk scoring model or upsell checklist. The published positions are surprise churn, signals living in conversations rather than usage data, process without signal, the named risk signals, failure modes by revenue type, the client relationship benchmarks of 10 percent concentration and three year tenure, client and end market quality, relationships with the firm rather than one employee, the opportunity signals, client fatigue, the SBI tenure story, the four expansion conversations, the 90 percent retention threshold, and the division of work between AI and people.

When this answer flips

If your revenue is mostly one-off projects, as an inference, the main risk flag is the absence of a next project, and the opportunity flags matter more than the churn flags.

If one client is above 10 percent of billings, treat concentration as a standing risk even while that account is growing.

And if your client records live in personal inboxes, start by collecting them, because nothing can be flagged that cannot be seen.

The short answer

Watch the same evidence for both. The retention essay says churn signals live in meetings, emails, deliverables and responsiveness, and names disengagement, value confusion, relevance drift, executive withdrawal and priority loss, with different failure modes for retainers, subscriptions, outsourcing, fractional roles and long projects. The account management essay names the openings: implicit requests, unspoken needs and receptive moments. Add the 2020 book checks on concentration, tenure and who holds the relationship. Put every flag on one list per account with an owner and a next step, match each to the right conversation, and let AI do the watching while people do the acting.