Will Superintelligence Disrupt Your Business? A Stress Test

Superintelligence will disrupt your business in proportion to how much of your revenue depends on scarce human cognition: analysis, drafting, coordination and routine judgment sold by the hour or by the deliverable. It will disrupt far less where customers pay for trust, relationships, accountability or assets that software cannot copy. You can learn where you stand without forecasting a date, by running a structural stress test on your own business model.

Key takeaways

• Exposure follows what customers actually pay for. If they pay for thinking time, you are more exposed than if they pay for accountability, access or ownership.
• The test needs no forecast. Assume that high-quality cognitive work becomes abundant and cheap, then ask what still protects your margin and your customer relationships.
• Trust, relationships and assets protect you only while someone tends them. Careless automation can damage all three.
• Score each activity on two questions: how much of its value is cognition, and how much depends on who delivers it. The pairing tells you whether to automate, redesign, defend or drop it.
• Finish with a short list of moves, each with one owner, and rerun the test on a schedule instead of treating it as a one-off strategy exercise.

Why run a stress test instead of waiting for a forecast?

Most debates about superintelligence stall on timing. One person expects rapid change, another expects slow change, and the leader is left with nothing to do. A stress test avoids the argument. A structural engineer does not predict the exact storm before checking a bridge. They apply a heavy load and see where it bends.

Your load is one assumption: any task that is mainly reading, writing, analyzing, planning or deciding on information can be done by software at high quality and low marginal cost. You do not have to believe this is near. You only have to ask which parts of your model still hold if it happens.

This is not a scenario exercise. Scenarios describe a future world and assign odds. The stress test uses facts you already have: your offers, your cost base, your contracts and your customer relationships. For a wider view of what shifts and what stays as models improve, see what changes and what stays in business when AI gets stronger. Here we stay inside your own model.

Step one: map revenue to the work that earns it

Start from revenue, not from the org chart. List every offer you sell. For each, write in plain words what the customer is really paying for. Be strict: a report is not an answer, but a conclusion the customer can defend to their own board is.

Then list the activities between a customer request and a delivered result, for example intake, research, drafting, quoting, scheduling, quality checks, delivery, follow-up, billing and exception handling. Keep it to one page. If it spills over, you are listing tasks, not activities.

Add one more column: who bears the consequences if the output is wrong. This column matters later, because accountability is one of the few things that does not become abundant when cognition does.

Step two: label each activity as cognition, trust, relationship or asset

Give every activity one primary label, and a second only if it truly needs it.

• Cognition: the value comes from processing information well, such as summarizing documents, preparing first drafts, comparing options, answering routine questions or reconciling records.
• Trust: the value comes from someone being accountable, such as signing off, certifying, advising where a mistake carries legal or reputational weight, or handling sensitive data.
• Relationship: the value comes from who is on the other side, such as renewing a long-standing account, resolving a complaint, making introductions or holding standing in a community.
• Asset: the value comes from something you own or control that is hard to replicate, such as permitted proprietary data, locations, licenses, equipment, distribution, exclusive contracts or a recognized brand.

Be honest about the cognition label. Leaders often call analytical work judgment because it feels senior. A useful check: if a capable newcomer with the right documents and a free day could do it, then the scarce part is the documents and the access, not the thinking. Ask whether the scarcity is in the mind or in the permission.

Also note the asymmetry. Trust, relationships and assets are not automatically safe. The technology that makes cognition cheap can erode them: a flood of convincing automated messages lowers the value of a personal reply, and a mistake repeated at scale damages trust faster than one person's slip. They protect you only if you decide deliberately how automation touches them.

Step three: score each activity and choose a posture

Ask two questions of each activity and answer high or low. First, how much of the delivered value is cognition that software could soon do well? Second, how much does the value depend on who delivers it, who is accountable or what you own? Then use this decision guide:

• High cognition, low dependence on who delivers: most exposed. Expect customers to demand this faster and cheaper from anyone. Automate it under human control and assume its price will fall. Your edge must come from the offer around this step.
• High cognition, high dependence on who delivers: transforming. Software does the heavy work and a named person stays accountable. Redesign the role so that person reviews and decides, and make that accountability visible to the customer.
• Low cognition, high dependence on who delivers: most protected. Invest in the people, relationships and assets involved, and use agents only to give them more time.
• Low cognition, low dependence on who delivers: commodity operations. Treat it as cost, not moat. Automate it if it is routine, outsource it if it is physical, or drop it.

Add two checks. First, ask where you would be uncomfortable if a customer saw how the work was produced. That line marks where trust is doing the work, and a person should own it or the method should be disclosed. Second, look at cost per unit of delivered value. If the cost of producing a unit falls while your price stays tied to effort, a competitor will pass the saving on to the customer before you do.

What does the test look like in practice?

Illustrative scenario, not a real client. A small firm sells monthly bookkeeping, year-end filings and occasional advisory calls. Customers say they pay for peace of mind about their numbers.

Labeling the activities gives this picture. Collecting documents, categorizing transactions and preparing statements are cognition with low dependence on who delivers. Reviewing exceptions and filing are cognition with strong accountability, because a named professional answers for them. Advising is trust and relationship. The firm's long-standing client contracts and its working relationships with banks and local authorities are assets.

The posture follows. Categorizing and preparing are most exposed, so the firm should expect rivals to offer them cheaply and should reduce its dependence on pricing by transaction volume. Review and filing are transforming: an agent prepares, the professional signs off. Advising is protected, so the firm should move partner time from preparation to advice.

The moves: reprice from effort toward scope and outcome, put an agent on document collection and first-pass categorization with a human reviewer, and write down which decisions only a named professional may make.

What do you do with the results, and how often do you repeat the test?

The test ends in a short list of moves, each with one owner:

• Pick the most exposed activity that is also frequent and well documented. It is the best first candidate for an agent, because you learn the most with the least risk.
• Define what a good result looks like before building anything, including who approves and what the agent may not do. The guide to success criteria for an AI agent pilot shows how.
• Decide where the gains go: margin, lower price or capacity for protected work. Keep memory, tools and prompts portable so the gains stay with you as models change, as covered in keeping AI gains in-house.
• Write a protected list of relationships, accountability points and assets, each with a named owner.
• Set a date to rerun the test.

Be Human's managed approach follows the same order. A first engagement defines one useful role, the memory and tools it needs, the approval boundaries and an observable success criterion. Managed operations then cover monitoring, recurring evaluations, permission review and controlled improvements, so the exposed work is done by an agent that a person can inspect while accountability stays with your people.

Rerun the test on a schedule, such as each planning cycle, and after events: a competitor changes how it prices, a customer asks how work is produced, a model you rely on changes materially, or a key person leaves. Each time, ask whether an activity moved between boxes and whether anything on the protected list weakened. A recognized risk framework, such as the NIST AI Risk Management Framework, is a useful checklist for the governance side. For preparation work beyond this test, read how to prepare your company for superintelligent AI today.

Two mistakes are common: trusting relationships you have not maintained, and automating everything out of alarm. The test exists to prevent both.

If you want help choosing where to begin, take the free Be Human AI diagnostic to find the first task worth handing to an agent.

Frequently asked questions

Will superintelligence disrupt a service business more than a product business? Not automatically. What matters is what customers pay for. A service firm that sells accountability and relationships can be better protected than a product company whose value is mostly information, while a service sold purely as hours of analysis is exposed.

Are relationships enough to protect my business? Only if they are active and tied to value the customer cannot easily get elsewhere. A relationship that exists on paper but has had no recent contact tends to weaken when a cheaper alternative appears. Treat relationships as assets that need an owner and regular attention.

Do I need to understand superintelligence technically to run this test? No. The test uses one assumption, that information work becomes cheap and good, and everything else is knowledge of your own business. You need to know what customers pay for, how work gets done and who is accountable when it goes wrong.

Who should run the stress test? A senior operator who knows how work actually gets done, with the finance lead and the person who owns key customer relationships in the room. Give one executive ownership of the result so the list of moves does not stall after the first meeting.

Be Human editorial guidance. Examples are illustrative unless explicitly attributed. Sources provide background; implementation choices depend on your workflow.