Once a company sees customer profitability clearly, it has to decide what to do with the truth.

There are four common moves: reprice, automate, redesign, or exit. The mistake is treating every bad-margin customer the same. Some customers should pay more. Some should be served differently. Some reveal product work that would make a whole segment profitable. Some should be allowed to churn.

Repricing is appropriate when the customer receives real value but the commercial model does not match cost. This is common in AI products where usage, inference, support, or workflow complexity grew beyond the original contract. Repricing works best when the company can explain the value metric and give the customer a fair path forward. It works poorly when the company simply discovers its own bad economics and pushes the pain onto the account.

Automation is appropriate when many customers create the same service burden. If onboarding, reporting, QA, training, or support repeats across accounts, the company may have a product opportunity. Automating that work can improve margin and customer experience at the same time. The key is repetition. Automating one customer's strange exception usually just hardens the exception.

Redesign is appropriate when the product creates the cost. Maybe the workflow requires too many steps. Maybe the AI output is hard to trust. Maybe permissions are confusing. Maybe pricing encourages wasteful usage. Maybe onboarding asks customers to understand the internal architecture. In those cases, customer profitability is a product critique.

Exit is appropriate when the customer is structurally wrong for the business. The account needs too much custom work, demands roadmap exceptions, resists pricing, creates support burden, and teaches little that applies elsewhere. Keeping that customer may protect short-term ARR while weakening the company.

The operator test: which action would make this segment more profitable if repeated ten times?

That question prevents account-by-account improvisation. Repricing one account may be necessary. Repricing a segment is strategy. Automating one support request may help. Automating a repeated customer journey changes the model. Exiting one account may feel tactical. Exiting a segment can restore focus.

Growth stage matters again. A high-growth company may delay exits while it learns. A low-growth company may need to act faster because bad customers consume scarce capacity. But even high-growth companies should label the subsidy. If a customer is being kept for learning, say so. If a customer is being kept for logo value, say so. Hidden subsidies become bad strategy.

The customer conversation should be honest. "We cannot support this custom workflow at the current price." "This use case belongs on a different package." "We are standardizing this capability." "We are not the right vendor for this operating model." Clarity is better than quiet resentment.

Internally, the decision should include sales, success, product, finance, and support. Each function sees a different part of the cost. Sales sees willingness to pay. Success sees adoption. Product sees roadmap fit. Finance sees margin. Support sees friction. The profitability decision needs all of it.

Customer profitability is not about becoming hostile to customers. It is about serving the right customers with an operating model that can compound. That sometimes means charging more. Sometimes it means building better product. Sometimes it means saying no.

The action should be matched to the cause. If the customer is profitable except for one repeated manual step, automate. If the customer is valuable but underpriced, reprice. If the customer exposes a broad product gap, redesign. If the customer keeps demanding exceptions that do not generalize, exit or contain the relationship.

Containment is sometimes the right middle path. Freeze custom scope. Move the account to a paid services tier. Stop accepting new bespoke commitments. Require executive approval for exceptions. The goal is not drama. The goal is to prevent one customer from quietly becoming the operating model.

Customer-facing teams need language for these moves before the crisis. If the first honest conversation happens at renewal, the company will sound reactive. Better to set expectations during onboarding, usage reviews, and expansion planning: which work is standard, which work is paid, which work is experimental, and which work is outside the product boundary.

That boundary protects both sides. Customers know what they are buying, and the company knows what it is agreeing to serve.

The company should also be careful with heroic saves. Saving a bad-fit customer can feel like customer obsession, especially near renewal. But if the save requires custom promises, senior intervention, or unpriced work that will continue next year, the renewal may simply defer the decision. Some saves are worth it. Some are just delayed churn with extra cost.

A cleaner process ranks accounts by repairability. Can this account become healthy through pricing? Through product work? Through automation? Through clearer scope? Through a different package? If the answer is no across the board, the honest move may be to stop investing in the relationship.

The best decision is usually the one that can be repeated without heroics. If the company would not want to make the same exception for ten similar accounts, it should pause before calling the move strategic.

Evidence note: this series uses SaaS gross-margin, retention, and AI cost context as background for customer portfolio decisions: https://www.saas-capital.com/research/private-saas-company-growth-rate-benchmarks/ and https://www.tanayj.com/p/the-gross-margin-debate-in-ai

This is part 9 of 10 in Customer Profitability in the AI Era.