Exponential Transformation: Who Gets a Say in How the Work Gets Done?

Sariio exists to put work preference insight in the hands of the people doing the work. Every person should have access to an open, structured language for their own contribution and a say in how the work gets done. That is the purpose of democratisation for Sariio.

In Peter Diamandis’s Six Ds, digitisation makes a capability easier to distribute, and democratisation describes what happens when more people can use it. For Sariio, democratisation is a deliberate choice about who should have access. Digitised preference data and AI make it practical to tailor reports for individuals, colleagues, teams and leaders across an organisation. Diamandis’s account of the Six Ds.

Salim Ismail, Francisco Palao and Michelle Lapierre set out the ExO Sprint in their book Exponential Transformation. The sprint is a ten-week process for applying the Exponential Organisation model. Peter H. Diamandis wrote the foreword. Wiley’s book listing. Diamandis’s foreword.

Sariio is the work preference layer. It draws on preference data to enable people to shape their own contribution and commit to doing their best work together.

What the Six Ds help leaders examine

Digitisation

Digitisation puts information into a form that software can store, search and reproduce. A searchable set of service records can be used in ways that a filing cabinet cannot. Someone can find similar cases, compare how they were resolved and share the relevant material with a colleague at another site. The important question is what the record represents. A closed case may mean the problem was solved, or simply that a queue was cleared. If those meanings are mixed together, the digital record can spread the ambiguity throughout the business.

There is a further distinction between putting documents online and making the underlying work available for analysis. A digital image of a form may preserve the evidence while leaving the fields impractical to compare. If the goal is to find which decisions are routinely referred upwards, the useful record includes who made each decision, what information was available and why the referral took place.

Deception

In Diamandis’s framework, deception is the period when compound improvement is hard to recognise because the absolute gains are still small. A change from 1,000 to 2,000 transactions looks modest next to a market that processes a million. The same proportional increase at a much larger scale can alter which suppliers or services are viable.

Early disappointment is ambiguous. A product may be improving while it struggles to meet the standard required for use. It may also fail because the approach does not address the problem. Before treating a weak pilot as a sign of future growth, identify the specific constraint and track whether it is easing. If the problem is the tool’s ability to handle rare cases, another demonstration using typical cases tells the business very little.

Disruption

A capability becomes disruptive when it changes the basis on which people choose between alternatives. A tool that helps a specialist complete a familiar task faster may improve the economics of the existing service. A tool that allows clients to complete the task independently can change what they purchase and when they need specialist advice. The same technical advance can have both effects in different parts of a market.

In personality and psychometric publishing, one convention Sariio challenges is the emphasis on producing a profile that an assessor interprets. Some instruments are designed for specialist use, and the evidence for any instrument depends on its intended purpose. An everyday conversation about how to work together calls for a language that people can understand and use directly. Read Sariio’s review of Annie Murphy Paul’s critique of personality testing.

That changes the purpose of the report. A work preference report enables a person to describe what helps them prepare, decide and contribute. Colleagues can compare preferences by mutual choice and use the differences they find to agree how they will work together. The decision remains with the people responsible for the work.

Demonetisation

Stanford’s 2025 AI Index reports that the price of inference at a fixed threshold on the MMLU benchmark fell from $20 to $0.07 per million tokens between November 2022 and October 2024. That was a reduction of more than 280-fold. The comparison holds the benchmark threshold constant. It does not equate the models’ performance on each task or measure the cost of employing them in a workflow. Stanford AI Index, research and development.

A department may pay much less to generate a report while spending more to verify the underlying data. The new report may still be worth producing, particularly if the department can examine cases that previously went unreviewed. The financial case then depends on the value of the additional investigation. A token price cannot settle that question. The useful denominator is the cost per acceptable result, with review and rework included.

Dematerialisation

A dedicated device or product may disappear as its function becomes part of a more widely available system. A phone can take a photograph that once required a separate camera. In an organisation, a dedicated reporting tool may disappear into the system where people already record and manage work. A separate login and export step can disappear with it.

Look at which responsibilities were attached to the product before retiring it. A regular report may have brought together people who checked the definitions and noticed changes in the underlying records. If the report becomes a self-service dashboard, decide who now watches for those changes and how other users will hear about a revision. The purpose of the old arrangement may survive the software that supported it.

Democratisation

A capability is democratised when a wider group can use it. A team that can query its own data may be able to investigate a problem before commissioning a report. Wider access can also reveal a gap between being able to generate an answer and having the expertise or authority to act. Access to a tool is easiest to count. The distribution of authority takes closer examination.

Sariio’s strategic intent is to democratise work preference insight. Specialist interpretation and the cost of producing individual reports could concentrate that insight among managers or people selected for a development programme. Digitised preference data and AI make it practical to tailor an explanation to the person reading it and the decision at hand. The choice to make that insight available across the organisation comes from Sariio’s purpose.

An individual can see their own preferences and decide what to share. Colleagues can compare reports by mutual choice and agree how to prepare for a decision. A team can use the same language to decide how to share responsibility. Leaders can use a view across teams to see where a strategy calls for a common standard and where teams need room to decide how to meet it.

The value of wider access depends on whether people can act on what they learn. If a team is expected to make decisions faster, its members need a say in what evidence they share, when they review it and which decisions they can make locally. That is how the work preference layer connects strategy to execution. The people responsible for delivering the strategy can shape the way they carry it out.

What workplace studies measured

In the published Quarterly Journal of Economics study by Erik Brynjolfsson, Danielle Li and Lindsey Raymond, researchers analysed the phased introduction of an AI assistant in technical customer support. Their data included 5,172 agents. Access increased issues resolved per hour by 15% on average. Gains varied by experience and skill. Under the studied workflow, agents could use the tool’s recommendations while handling live chats. The study measured a unit of service delivered, which makes the finding more useful for a performance discussion than a count of suggestions generated. It remains an estimate from a particular type of work and deployment. Generative AI at Work.

A field experiment with 758 consultants led by Fabrizio Dell’Acqua examined performance on tasks with and without AI. On a task the researchers selected because it lay beyond the system’s capabilities, consultants using AI were 19% less likely to produce a correct solution, as reported in the authors’ abstract. The same study found gains on tasks within the system’s capabilities. The practical implication is that adjacent assignments can need different rules for using the same tool. A successful drafting task provides little evidence that the tool can also diagnose an inconsistency in the source material. Navigating the Jagged Technological Frontier.

The support study and the consulting experiment answer different questions. One observes a rollout in an operating service and measures resolutions per hour. The other uses experimental assignments to compare performance on particular tasks. A leader can use both when designing an evaluation: measure what changes in the live work and test difficult cases deliberately, even if those cases occur too rarely to move the average in the first month.

Why the gain may take time to appear

Brynjolfsson, Daniel Rock and Chad Syverson’s Productivity J-Curve examines the investments that organisations make to use a general-purpose technology. New processes, skills and other intangible assets take time to build. Their value can be poorly captured in conventional productivity statistics. Early costs can therefore depress measured productivity, with later use of the assets raising it. The paper provides a model and historical analysis of complementary investment. A firm still needs to establish whether its own investment is creating a capability that will be used. The Productivity J-Curve.

The Six Ds look at how the capability and its market can change. The J-Curve directs attention to the firm’s investment in making the capability productive. The workplace studies then provide examples of how to test results at the level of a task or service. The three perspectives are most useful when kept close to their respective measures. A falling input price is a reason to reconsider a workflow. An increased cost during transition is worth investigating if it reflects deliberate investment. Neither figure establishes that customers now receive a better service.

Suppose a firm automates the first pass over a set of applications. If the same reviewers need to check a larger number of recommendations with fewer minutes available per case, the amount of time spent generating recommendations tells only part of the story. An evaluation can compare completed cases, the length of the review queue and the rate of decisions returned for correction. It can also track whether rare cases receive more attention because reviewers now have time available, or less attention because the standard output looks convincing. These are proposed measures for a workflow, not results from the studies.

Who decides how the redesigned work runs?

When a workflow changes, people often acquire new responsibilities without a formal change in role. A colleague who used to prepare evidence may now be expected to verify the evidence that a system retrieves. The work calls for different judgement and may require access to original records that were previously someone else’s responsibility. A new tool can make that change easy to overlook because the request arrives on the same screen as before.

At the individual level, ask what information a person needs to stand behind a decision and whether they can obtain it. Between colleagues, agree what a handover must contain so the next person can check it without repeating the work. At team level, decide which cases are safe to process through the new route and who can stop an unusual case. The organisation can set common requirements for assurance and review the costs that move between departments. Local discretion is easier to defend when its boundaries are clear and the standard for completed work is shared.

What to measure before expanding a rollout

Choose a unit of work with a clear end. An application is complete when a decision can be defended, not when the draft recommendation appears. A service issue is complete when it is resolved, including any follow-up needed to confirm the result. Record how long completion takes, what quality standard applies and who does the work at each stage. This makes it possible to see where time has been saved and where effort has moved.

Use comparable cases, and examine the difficult ones separately. A pilot with a stable mix of work will be easier to interpret than a comparison between a quiet month and a busy one. Include review effort and errors that escape review. Ask the people who do the work which checks still carry information and which checks simply replicate an earlier step. If teams make different choices, keep enough detail to tell which approach suited which work.

Expand when the evidence supports a better completed result at a cost the organisation can sustain. If the pilot speeds up preparation and lengthens the queue for approval, the next investment may belong in review capacity or in a clearer rule for who can approve routine cases. Knowing where work went after it became cheaper to start is part of knowing what the transformation achieved.

The work preference layer in exponential transformation

Democratisation is a choice about who can shape the work. When people across an organisation can draw on their own work preference insight, they can take part in decisions about how the strategy becomes everyday work. Digitisation and AI make personalised reporting practical at scale. Sariio uses that capability to widen participation in decisions about the work.

Say the strategy calls for faster decisions across several sites. An individual can explain what evidence they need to take responsibility for a decision. Colleagues can agree how to exchange that evidence. A team can agree which cases to decide locally and when to ask for another view. At organisation level, leaders can see where a common review protects quality and where teams can safely exercise discretion.

The work preference layer links strategy to execution through the agency of the people doing the work. We do our best work when we can ‘play to our preferences’.

Frequently asked questions

What are the Six Ds of Exponential Transformation?

The Six Ds are a framework associated with Peter Diamandis. They are digitisation, deception, disruption, demonetisation, dematerialisation and democratisation. They describe ways a digital capability can improve, fall in cost and reach more people. An organisation can use the framework to examine how a market or workflow might change. The timing and order depend on the context. Diamandis’s Six Ds.

What does exponential transformation mean for business performance?

A digital capability may become cheaper or more widely available fast enough to alter how a business delivers a service. Performance depends on what happens after adoption. A sound evaluation checks the quality, time and full cost of completed work, including review and rework. Faster production at one stage may shift effort to another stage.

How much did the cost of AI inference fall in Stanford’s comparison?

Stanford’s 2025 AI Index recorded a fall from $20 to $0.07 per million tokens between November 2022 and October 2024 for models meeting a fixed MMLU benchmark threshold. That represents a reduction of more than 280-fold in the reported inference price. The comparison does not include the cost of integrating models into business processes. Stanford AI Index.

How does the Productivity J-Curve relate to the Six Ds?

The Six Ds describe changes in a capability’s availability and economics. The Productivity J-Curve explains how investment in skills and new processes can precede visible productivity gains. Used together, they prompt two separate enquiries into whether a capability is becoming viable and what it will cost to make productive use of that capability. Productivity J-Curve research.

How can leaders tell whether AI improves a particular workflow?

Compare a completed unit of work before and after the change. Check whether the result meets the same quality standard, where delays accumulate and who spends time on verification. Test difficult cases separately. That gives the organisation a basis for deciding which uses to expand and which need more evidence.

How does Sariio apply disruption and democratisation?

Sariio uses work preference reports to create a shared language that people can use directly. Personalised reporting makes that language available at individual, interpersonal, team and organisation levels. People can use it to shape their contribution and agree how to meet a shared objective.