Skip to content
finalstate.aiDifferentiated Value: The Win Everyone Can Buy Is Not an Edge
118Atlas

118AFTER ADOPTION

Differentiated Value: The Win Everyone Can Buy Is Not an Edge

A workflow gain can save a firm money without giving it an advantage over rivals. An illustrative support ledger separates realized savings, costs and competitive claims; adoption and productivity statistics do not supply its amounts.

A support queue and time log lead to an unresolved savings entry.Completed support tickets release time. A financial entry remains a question until the team can show realized economics. The visual does not convert a productivity percentage into dollars.Hours$?Savings?
Illustrative support workflow. The later ledger compares the same case volume and service quality over 90 days, in thousands of US dollars. No survey or field-study coefficient is converted into these amounts.

THREE CLAIMS

A useful gain can remain common

Support tickets, a financial receipt and a rival comparison represent three different claims.The workflow log has completed tasks. Its financial receipt and rival comparison remain unresolved. Separate questions prevent an operating metric from being treated as earnings or a relative advantage.Workflow$?Accounts?Rivals

Operating evidence concerns time, quality, throughput or risk in a workflow. Enterprise impact requires a contribution to the accounts after costs and offsets. A competitive advantage requires comparison with relevant rivals over a stated period. A firm can retain positive savings even when rivals can do the same.

DIFFERENT SURVEY QUESTIONS

Adoption and EBIT are different survey questions

Three separate survey entries show 78 percent AI adoption, 71 percent regular genAI use and 17 percent reporting at least 5 percent of enterprise EBIT attributable to genAI.The entries have no connecting funnel or subtraction. The full survey has 1,491 participants; the denominator for the EBIT question is not established. The result does not mean that everyone below its reporting threshold received zero value.RespondentsAI adoption≥1 function78%Regular genAI≥1 function71%GenAI ≥5%of enterprise EBIT17%EBIT base unstated
Historical respondent self-report, weighted by national GDP. “Adoption” was deliberately undefined. Full survey n=1,491; this is not an established EBIT-question denominator. Report pp. 14–16, 21 footnote 7, and 24: https://www.mckinsey.com/~/media/mckinsey/business%20functions/quantumblack/our%20insights/the%20state%20of%20ai/2025/the-state-of-ai-how-organizations-are-rewiring-to-capture-value_final.pdf

McKinsey’s March 12, 2025 report draws on July 16–31, 2024 fieldwork with 1,491 participants. It reports 78% AI adoption in at least one function and 71% regular genAI use. Separately, 17% report that at least 5% of enterprise EBIT in the preceding twelve months was attributable to genAI. The EBIT question’s base is not stated in the cited passage. These figures cannot be subtracted to estimate failure or drawn as a conversion funnel.

The survey does not tell you where a saving went

Failing to report the EBIT threshold is not a report of zero value. Costs, limited use, timing and measurement could matter; customer pass-through is another hypothesis to investigate. The report also associates workflow redesign with reported impact. That association does not establish that redesign caused the gain or is necessary in every deployment.

ONE OPERATING GAIN

Support productivity is one measured outcome

Fixed bars show a 15.2 percent overall increase and a 36 percent increase for the lowest prior-skill quintile in issues resolved per hour.Both bars use the same zero-to-40-percent scale. The subgroup is based on prior performance, not simply novice status. No arbitrary bar is drawn for experienced workers, and no study result is converted to savings or competitive advantage.Issues resolved per hourOverallLowest priorskill quintile+15.2%+36%010203040Percent increase
Final QJE edition, 140(2), 889–942, published February 4, 2025. §V.A/Table II, column 3: +0.301 resolutions/hour (SE 0.0329), 15.2% relative to the pretreatment mean. §V.B.1/Figure III.A: 36% for the lowest prior-skill quintile, not all novices. Each increase is relative to the relevant pretreatment baseline. Point estimates shown; no uncertainty interval was reconstructed for the subgroup. https://doi.org/10.1093/qje/qjae044

The published 2025 study by Brynjolfsson, Li and Raymond covers 5,172 support agents at one firm. Its adjusted estimate is about 15% more issues resolved per hour; the lowest prior-skill quintile gained 36%. Skill here combines earlier speed, resolution and customer-satisfaction measures. It is distinct from tenure. The staggered rollout was not randomly assigned, and these estimates do not establish customer prices, retained savings or a durable competitive position.

Better operations and a distinctive position are separate claims

In What Is Strategy?, Michael Porter distinguishes improving comparable activities from configuring activities differently. His 1996 argument links positioning to tradeoffs and fit across activities. Applying that distinction here: an AI-assisted support workflow can be worth funding even if rivals can reproduce its improvement.

ONE PASS-THROUGH PATH

A rebate needs an agreement

A rival offer enters a renewal discussion, a three-thousand-dollar rebate is agreed, and a credit reaches the customer.The contract changes before the credit moves. A fixed three-thousand-dollar customer rebate is recorded once in the later ledger; no market-wide pass-through rate is asserted.RenewalOffer$3kFirmCustomer$3k
Illustrative transfer, in thousands of US dollars over 90 days. The $3,000 customer rebate is one ledger deduction. It is not inferred from the survey or productivity study.

In the illustrative support case, a customer cites a rival’s offer during renewal and obtains a $3,000 rebate over the same 90 days. The revised contract and credit establish this transfer. This is one possible bargaining path; savings can also affect wages, service or prices through other arrangements. Diffusion and bargaining are conditions to investigate, not a universal sequence.

CANDIDATE ADVANTAGES

Useful context does not settle who keeps the gain

Customer records improve a support workflow and receive a validated update, while portability to a rival remains a separate rights question.A record archive feeds the workflow, a checked resolution returns to it, and a questioned export route reaches another provider. The drawing treats learning and access as mechanisms to test, not proof of exclusive capture.Case historyWorkflowRival?Portability

A customer's case history may make an assistant more useful, and validated resolutions may improve that history. Test those mechanisms. Then inspect access rights, portability, renewal terms and bargaining: the customer might take the records elsewhere or claim the benefit in a lower price. Complements and reinforcement are candidates for differentiation, not universal requirements for positive savings. Accountable execution alone creates no exclusive right.

A 90-DAY RECONCILIATION

Recognize savings before subtracting costs

A ledger subtracts fifteen thousand dollars of non-overlapping deductions from realized savings and displays negative, zero and positive alternatives.The main case realizes twenty thousand dollars and deducts six for setup, two for extra review, four for vendor fees and three for a rebate, leaving five thousand. Alternative realized savings of zero and fifteen thousand give net minus fifteen thousand and zero. Delayed benefits are not counted in any current-period result.$k / 90 daysDeductionsSetup6Review2Vendor4Rebate315AlternativesSavedNet200−15150+5Subtract once
Illustrative incremental contribution, not audited EBIT or a study estimate. All amounts are thousands of US dollars over 90 days, relative to the same without-AI case volume and quality. Costs are expensed in this horizon and categories do not overlap; other revenue and costs are held constant. Avoided overtime must be evidenced; hours merely redeployed are not credited as expense savings. Future benefits remain outside the period until realized.

For this illustration, the without-AI baseline is $100,000 of overtime expense for the same 90-day workload and service quality. Verified expense of $80,000 realizes $20,000 of gross savings. Subtract mutually exclusive amounts: $6,000 internal setup, $2,000 additional review, $4,000 vendor fees and the $3,000 customer rebate. The net contribution is +$5,000. Vendor fees appear once; there is no second supplier-capture deduction. With the same $15,000 deductions, realized savings of $15,000 break even and $0 produce a −$15,000 result.

TEST THE RELATIVE CLAIM

Keep the saving; test the edge

A known five-thousand-dollar contribution sits beside an unresolved rival result, while a separate calendar leaves later benefits pending.The same-period comparison does not assign zero to the unknown rival. A separate future calendar keeps delayed benefits out of current earnings. Positive, zero or negative current value and timing remain different dimensions.Firm+$5kRival?90 daysLater?
Illustrative operator check. This firm’s +$5,000 is known within the example; the rival result and benefits beyond 90 days remain unresolved. A common improvement can still be worth taking.

Keep the ledger’s sign separate from its timing. A benefit expected after day 90 is pending, not a current saving and not a fourth kind of profit. Compare the realized contribution with a relevant rival on the same workload, quality and horizon before claiming an edge. Then examine how contracts, bargaining and imitation could change the result. Return to 004 with the evidence that should change the next funding decision.

Read the transcript

01 · AFTER ADOPTION

Begin with an illustrative support team after it has selected an AI tool. The queue moves faster and the time log records fewer overtime hours. That is an operating observation. The savings entry still needs evidence that spending fell or profitable work increased. Released capacity may have value even when it has not yet changed the accounts. This example will use the same case volume and service quality over ninety days. Its dollar amounts are invented to explain a reconciliation, not calculated from the studies that follow.

02 · THREE CLAIMS

Keep three claims separate. The workflow log records time, quality, throughput, or risk. The financial receipt asks whether the gain reaches the accounts after costs. The rival comparison asks whether the firm is better placed than a relevant competitor over a stated period. A team can save money using the same tool as everyone else. That saving may be worth keeping even when it supplies no distinctive position.

03 · DIFFERENT SURVEY QUESTIONS

McKinsey’s March 2025 report uses different questions. In fieldwork conducted in July 2024, seventy-eight percent reported AI adoption in at least one business function. Seventy-one percent reported regular generative-AI use. Separately, seventeen percent said at least five percent of enterprise earnings before interest and tax in the preceding twelve months was attributable to generative AI. The full survey had one thousand four hundred and ninety-one participants. The cited earnings passage does not state its own question base. These are self-reports with different scopes, not stages in a conversion funnel.

04 · EVIDENCE STOPS HERE

Someone who does not report its earnings threshold has not necessarily reported zero value. Costs, limited use, timing, or measurement could matter. Savings reaching customers is another possibility, but the survey does not establish that path. Its association between workflow redesign and reported impact also does not prove that redesign caused a gain, or that every deployment requires it. A local ledger must establish what happened in the particular workflow.

05 · ONE OPERATING GAIN

The final 2025 Quarterly Journal of Economics article by Brynjolfsson, Li, and Raymond covers five thousand one hundred and seventy-two support agents at one firm. The adjusted overall estimate was about fifteen percent more issues resolved per hour. The lowest prior-skill quintile gained thirty-six percent. That group was defined using earlier speed, resolution, and customer satisfaction; it is not simply the newest agents. Managers selected the timing of the staggered rollout, rather than assigning it randomly. The result does not tell us what customers paid or how much value the firm retained.

06 · PORTER'S DISTINCTION

Michael Porter’s 1996 article distinguishes improvements in comparable activities from a different configuration of activities. His argument about strategy turns on positioning, tradeoffs, and how activities fit together. Applied to this support team, faster work may pay for the investment while remaining reproducible elsewhere. A strategic claim needs evidence about the firm’s position relative to rivals. The tool’s presence, or even a positive savings entry, does not establish that comparison.

07 · ONE PASS-THROUGH PATH

Consider one possible path in the illustrative support business. A customer brings a rival's offer into the renewal discussion. The firm agrees to a three-thousand-dollar rebate over the same ninety days and issues the credit. Those records establish a transfer to the customer. Comparable technology alone would not show that it happened. Other arrangements may move value through wages, service commitments, or prices without this sequence. The ledger will count this rebate once, as a reduction in revenue.

08 · Advertisement · Twain GTM Agents

A good outbound message should know why it is being sent. Twain GTM agents research accounts, qualify contacts, and turn live signals into relevant outreach. They work with your CRM, personas, and GTM stack, so the message is not just personalized. It is backed by a specific reason to start the conversation.

09 · CANDIDATE ADVANTAGES

The support assistant may work better with a customer's case history. Validated resolutions could improve that history as the workflow is used. Test that improvement, then inspect the access and renewal terms. Can the customer export the same records to another provider? Can it bargain away the benefit? Context may make the workflow useful without making its value exclusive to this firm. Reinforcement is one possible source of differentiation. Ordinary positive savings do not require it, and doing the work responsibly does not create an exclusive right.

10 · A 90-DAY RECONCILIATION

Use one baseline, horizon, and monetary unit. In this illustration, the same workload and service quality would require one hundred thousand dollars of overtime over ninety days without AI. Verified expense of eighty thousand realizes twenty thousand in gross savings. Subtract six thousand for internal setup, two thousand for additional review, four thousand for vendor fees, and the three-thousand-dollar customer rebate. These categories do not overlap. Other revenue and costs are held constant. The net contribution is five thousand dollars. Do not subtract supplier capture again after counting vendor fees. With the same fifteen thousand in deductions, fifteen thousand of realized savings breaks even; no realized savings gives a fifteen-thousand-dollar loss. Hours redeployed are not automatically avoided expense.

11 · TEST THE RELATIVE CLAIM

The ledger's sign and its timing are different questions. A benefit expected after day ninety remains pending; it cannot fill the current savings entry. In the example, the firm has a five-thousand-dollar contribution for this period. The comparable rival result is still unknown. Check the same workload, quality, and horizon before claiming a relative advantage. Then examine what contracts, bargaining, or imitation could change. The saving can be worth taking while that investigation remains open. Bring the result back to the deployment-selection decision.

01 / 11 · AFTER ADOPTION0:00 / 6:45