The short answer. Buyers already value subscription and services revenue differently. But neither category shows how much revenue depends on work that AI can now perform. That work can exist in both categories.
Salesforce's filings show why the question matters. Professional services revenue fell last quarter while total revenue grew almost 11 percent. The cost of delivering those services also increased. The public numbers do not prove that AI caused the change.
The Revenue Substitution Score asks three questions about each revenue line: Can AI do the work? Who benefits when the work costs less? Is the work different for each customer? The answers place the revenue in one of three bands: durable, capped, or declining. That classification determines what management should assume in the FY27 plan.
Table of Contents
Research Grounding
Forrester published its AI Disruption Model on August 19. It covers more than 200 technology and services markets across 17 categories. The model considers how easily AI can perform the work, how much human labor the work requires, and how difficult it is for customers to switch providers.
Forrester expects three categories to benefit broadly from AI growth: infrastructure, data and AI, and identity, access, and network security. It expects more pressure on labor-intensive markets such as technology implementation, software development, and transformation services.
Recent company results show why this issue belongs in the operating plan. They do not prove that AI caused the changes.
Salesforce, quarter ended July 31, reported August 26. Professional services and other revenue fell from $546 million to $525 million. Total revenue grew 10.8 percent to $11.345 billion.
Accenture, quarter ended May 31, reported June 18. Revenue grew 6 percent to $18.7 billion. New bookings of $19.3 billion were 2 percent lower than the prior year.
Gartner's July 27 forecast, at market level. IT services spending is expected to grow 5.3 percent in 2026, compared with 14.2 percent growth for total IT spending.
Salesforce's professional services business lost $103 million at the gross-profit level, compared with a $51 million loss in the same quarter last year. Revenue fell 3.8 percent while the cost of delivering the services rose 5.2 percent to $628 million.
Several things could explain this pattern. Lower-cost work may be leaving while more expensive work remains. Cost reductions may be taking longer than the revenue decline. Staffing levels or project mix may also have changed. One quarter cannot tell us which explanation is correct. That is why management should review at least eight quarters of company data.
If lower-cost work is leaving, the filings do not show where it went. AI may be performing some of it. An implementation partner may also be taking the work. Either outcome changes the economics of the delivery business and may require a change to the plan.
The PE Translation
Buyers already separate subscription revenue from professional services revenue and value them differently. Subscription revenue is usually valued more highly because it is recurring, predictable, and easier to scale. Services revenue is usually valued less because it depends more directly on people and often carries a lower margin.
That distinction matters when buyers evaluate ARR, net revenue retention, gross margin, and overall revenue quality. It is the right place to start.
But it does not answer the next question: How much of the work behind each revenue line can AI perform, and who receives the financial benefit?
Implementation, configuration, integration, testing, data migration, training, and managed delivery all require people to complete work. Some of that work is sold separately as professional services. Some supports subscription revenue, bundled implementations, renewals, managed offerings, or customer expansion.
When AI reduces the hours needed to complete the work, three things can happen:
The vendor keeps the savings as margin. The price holds while the delivery cost falls. Revenue stays the same and services margin improves.
The value moves into the product. The company turns the capability into software. It may capture the benefit through higher subscription revenue, better retention, more customer expansion, or stronger margins.
The value leaves the company. The customer takes the work in-house, an implementation partner takes it, or the work is no longer billable at all. In each case the vendor loses the revenue and gains nothing in the product.
The first two keep the value inside the company. The third does not.
The first outcome is the easiest to achieve. It improves margin without changing the product, and it holds until the customer sees the work take less time and asks for some of the savings at the next renewal. The second outcome costs more because it requires product investment. It is the one that survives the next negotiation.
A decline in services revenue can be positive if customers go live faster, adopt more of the product, expand their subscriptions, renew at higher rates, or become less expensive to serve. It is less attractive if the services revenue disappears without improving any of those results.
Buyers already discount services revenue. A seller can make a stronger case by showing which services revenue remains valuable, where AI improves product economics, and whether any recurring revenue depends on work customers may stop paying for.
Issue 8, the METER Test, asked whether a company can change what it charges for. Issue 10 asked how adding AI to the product could affect valuation. This issue asks a different question: As AI changes how work is delivered, which revenue remains durable and who keeps the financial benefit?
A CFO may already track how often customers buy implementation services with the software. That number shows how much services revenue the company sells. It does not show how much of the work AI can perform or whether the benefit will appear in ARR, retention, or margin.
Operator Experience
Enterprise software often requires substantial implementation work, whether it is delivered as SaaS, hosted software, or installed on the customer's systems. The services component becomes significant when deployment requires data integration, migration, customer-specific business rules, workflow configuration, or changes to operating processes.
I saw this directly with the analytics and AI platform at Hoonuit, which later became part of PowerSchool. Each deployment required services work to connect source systems, map and validate data, configure customer requirements, and fit the platform to the way each organization operated. The same pattern appears across analytics, ERP, PLM, human capital systems, financial software, and many industry-specific products.
Some implementation tasks are becoming easier to automate because the instructions are clear and the results can be tested. Examples include:
mapping fields from a customer's old system into the new platform
writing data transformation rules
building a connector to a documented API
creating test data and test cases
producing configuration and training documents
AI agents can assist with or perform more of this work when they can access the required data and when the company can detect errors before the results reach the customer.
Other work remains harder to automate. Someone still has to understand how the customer operates, decide which old processes should change, resolve disagreements, and determine which exceptions are worth supporting. This work requires judgment and company knowledge that may not be documented.
Custom work is no longer protected from automation simply because it is different for every customer. The technical portion of a custom implementation may change quickly. The work that requires business judgment and customer decisions is more durable.
Engineering and delivery teams often see the change before finance does. A reduction in implementation hours per new customer may look like a productivity improvement. It could also mean that customers need less billable work. Both may be good for the customer, but they affect the company's revenue and margin differently.
Management should first make the comparison fair. Use median internal implementation hours for similar customers. Exclude work delivered by partners. Account for customer size, product mix, and the number of modules purchased.
Then review these measures across eight quarters:
internal implementation hours for similar new customers
services revenue and margin for those customers
the percentage of new sales that required custom integration work
the time required for customers to go live and begin using the product
renewal rates, customer expansion, and net revenue retention
Together, these measures show whether lower delivery effort is improving the product and its margins or simply reducing billable revenue.
Board Question
Which of our revenue lines depends on work that AI can now perform, and what does the FY27 plan assume will happen to that revenue?
The subscription and services split is the starting point. It does not show who benefits when AI reduces the work required.
The Revenue Substitution Score
Review each important revenue line using three questions. Answer them in order. Do not combine them using untested weights.
1. Substitution: Can AI do the work?
Does the task have clear instructions, a known correct result, and a reliable way to check the answer? If so, AI may be able to perform a meaningful part of the work. If the task depends on judgment, negotiation, or knowledge that has not been documented, it is more durable.
2. Revenue capture: Who benefits when the work costs less?
Determine whether the vendor, the customer, or an implementation partner receives the financial benefit.
A customer may remain on the platform because leaving would require rebuilding configurations, integrations, and data. That protects the subscription relationship, but it does not automatically protect services revenue. A customer can use the software for many years while purchasing implementation services only once.
The vendor can keep the benefit in two ways. It can hold the price while the delivery cost falls, which leaves revenue intact and improves services margin. Or it can build the capability into the product, where the benefit may appear as stronger ARR growth, better retention, faster adoption, more customer expansion, or higher margins.
The first way holds only until the next negotiation. The second requires product investment and lasts.
If the value leaves the company, the vendor loses the revenue and gains neither advantage.
3. Customization: Is the work different for each customer?
Work that is nearly identical for every customer has often already been standardized or automated. Custom work needs a closer review.
Separate custom work into two parts. The technical part includes mapping, data transformation, testing, and building connectors to documented systems. The organizational part includes understanding how the customer operates, resolving disagreements, and deciding which processes should change. The technical work is likely to change first.
The Revenue Substitution Score in one screen
SUBSTITUTION: Can AI do the work? If not, the line is durable, and the review stops here.
REVENUE CAPTURE: Who benefits when the work costs less? The vendor can hold the price and keep the savings as margin, or build the capability into the product. If the value leaves the company, services revenue falls with no product gain.
CUSTOMIZATION: Is the work different for each customer? If so, separate the technical work from the work that requires business judgment.
The answers place each revenue line in one of three bands.
Band | What to assume in the plan | Evidence |
|---|---|---|
Durable | Keep the planned growth rate and monitor for change. | Customers keep buying the work at the same rate and the price holds. Hours per similar customer may fall. If the price holds, that shows up as margin. |
Capped | Keep the current revenue, but assume no further growth through the hold period. | The price is holding, but the share of new customers buying the work is flat or slipping. Some customers have begun asking for some of the savings at renewal. |
Declining | Plan for the revenue to decline. Assign an owner and deadline for replacing it. | The share of customers buying the work, the price, or bookings are falling. Customers are taking the work in-house. |
The information request is straightforward. Ask for eight quarters of implementation hours, services revenue, and delivery margin for similar customer groups. Review those figures alongside time to go live, ARR growth, customer expansion, and net revenue retention.
Three Decisions
1. Review each revenue line before approving the FY27 plan
The CFO and head of delivery should create one table listing the company's important revenue lines. For each line, answer the three questions, assign a band, and record the evidence. Use this table alongside the existing subscription and services analysis to show how AI may change the economics within each category.
2. Track implementation hours and explain why they are changing
The technology leader should report raw implementation hours and the adjusted number for similar customers each quarter. Review both numbers with services margin, time to go live, ARR growth, customer expansion, and net revenue retention.
Management should state whether the reduction in hours reflects better productivity, lower demand for services, or both. It should also show whether the benefit appears in margin, customer adoption, expansion, or another financial result.
3. Decide where the value from lower delivery costs will go
Management should decide which of the three outcomes is most likely: the vendor keeps the savings as margin, the value moves into the product, or the value leaves the company. Assign an owner to the response and reflect the expected result in the operating plan.
The person responsible for integrations should separate technical implementation work from work that requires customer and business judgment. Evaluate the technical work for productization and include the cost of building and maintaining connections to customers' existing systems.
One Number for the Next Operating Review
$103 million. That was Salesforce's gross loss on professional services for the quarter ended July 31. The loss was $51 million in the same quarter last year. Revenue from the line declined 3.8 percent while its delivery cost increased 5.2 percent.
A services business that is shrinking while becoming less profitable requires investigation. Cost reductions may simply be taking longer than the revenue decline, but management should confirm that with evidence.
For a portfolio company, compare the same revenue and margin trend across eight quarters with implementation hours, time to go live, ARR growth, customer expansion, and net revenue retention. The goal is to determine whether delivery efficiency is creating product value or only removing services revenue.
Board Takeaway
Buyers already distinguish subscription revenue from services revenue. The next step is to identify which revenue depends on work AI can perform and determine who receives the resulting value.
Review the important revenue lines before approving the FY27 plan. Classify each as durable, capped, or declining, and connect the result to ARR, net revenue retention, gross margin, and the FY27 plan.
Portco Brief is a weekly briefing for PE operating partners and portfolio company executives focused on technology, AI, and value creation. If this was forwarded to you, subscribe at portcobrief.com.
Sources
Forrester, "Forrester Introduces AI Disruption Model To Assess AI's Impact On Technology And Service Markets", August 19, 2026.
Salesforce, "Salesforce Announces Second Quarter Fiscal 2027 Results", August 26, 2026.
Accenture, "Accenture Reports Third-Quarter Fiscal 2026 Results", June 18, 2026.
Gartner, "Gartner Forecasts Worldwide IT Spending to Grow 14.2% in 2026, Totaling $6.37 Trillion", July 27, 2026.