Skip to content

Start With What the System Needs Contributed

Brian Lambert, PhD
· 22 min read
Start With What the System Needs Contributed

Most work redesign starts with a list of tasks and a question: which of these can we automate? We think that is the wrong first question. Start with the beneficiary and the outcome. Derive the contribution the work system needs. Only then decide what a person does, what an agent does, and what both do together. Some tasks will not be allocated at all, because nothing needs them.

This article walks through that outside-in method on one workflow, shows where capability work belongs in it, and explains why we diagnose before anyone builds a course.

TL;DR

  • A task list describes how work is done today, handoffs and rework included. Automating it preserves the operating model.
  • Outside-in redesign starts from the beneficiary and the outcome, derives the required contribution, then allocates it to human, agent or both.
  • Capability need appears only after that allocation. Many gaps come from the environment, not the person, so check causes before building anything.
  • Use your own source material to decide: strategy, frontline friction, quality findings and expert knowledge.
  • Some tasks disappear instead of being automated. That is the strongest result.

Table of Contents

Cognistry
Diagnose Capability Before Building
 
Cognistry uses your organization's evidence to diagnose what capability work requires and whether learning is the right response.
Explore Cognistry

The spreadsheet everyone builds

You know this spreadsheet. Tasks run down the side. Across the top sit columns for automation potential: repeatable, rules-based, high volume, data-rich. Teams score each row, sort, and pick the top of the list.

It is a respectable tool. It forces a team to write down what people actually do. It surfaces effort nobody had counted. It gives a steering group something concrete to approve. For a quick efficiency pass on a stable process, it works.

But look at what it takes as given. Every row is a task that exists today. The spreadsheet can tell you which tasks are easy to automate. It cannot tell you whether the task should exist.

The blind spot: tasks that exist only because of old handoffs

A great deal of work is scaffolding. It exists because one team could not see another team’s system. Because an approval was added after one bad month. Because a form was never fixed upstream, so someone corrects it downstream.

Three patterns show up again and again:

  • Handoff tasks. Re-keying data from one system into another. Forwarding a case to a team that applies a standard rule.
  • Approval tasks. A sign-off that rarely changes the decision but adds days.
  • Rework tasks. Checking and correcting output that was wrong because the first step lacked the right information.

Score these on an automation spreadsheet and they rank high. They are repetitive and rules-based. Automating them feels like a win. But you have then built a faster version of a process that never needed to exist. The operating model survives, now with software holding it up.

That is the blind spot. Starting from today’s task list preserves the operating model that AI should make obsolete.

What the system needs contributed

The phrase sounds abstract, so make it plain. Any piece of work is a system: people, technology and structure interacting to produce something for someone. The Canada Energy Regulator describes workplace performance as emerging from interacting people, technology and organizational structures, with their interdependencies forming the context in which decisions and actions happen.

Researchers on work systems define one as human participants and/or machines performing work, using information, technology and other resources, to produce products or services for internal or external customers. The same work lists nine elements: customers, products and services, processes and activities, participants, information, technologies, environment, infrastructure and strategies, and says they should be aligned.

If you want a simple version of the three parts, use these:

  1. What goes in. The request, the information, the materials.
  2. What transforms it. The people, tools and steps doing the work.
  3. What comes out. The product or service the beneficiary receives.

Task lists live entirely in the second part. Outside-in redesign starts in the third. What does the beneficiary need to receive? What must be true for that to happen? Which of those truths needs a contribution, a judgment, a decision, a piece of information, an action, from someone or something?

That required contribution is the unit of design. A task is just one historical way of supplying it.

The outside-in method in four moves

  1. Name the beneficiary and the outcome. Who receives the result, and what does a good result look like to them? Write it as a sentence a customer or internal client would recognize.
  2. Derive the required contributions. Working backward from the outcome, list what must be supplied: information, decisions, actions, assurance. Ignore who does them today.
  3. Allocate each contribution. To a person, to an agent, or to both. Judgment under ambiguity, accountability and exceptions lean toward people. Assembling information and applying settled rules lean toward agents. Many contributions need both, with a clear boundary.
  4. Check what is left. Compare against today’s task list. Tasks that supply no required contribution are candidates to remove, not to automate.

Keep this to one workflow at a time. Redesigning how a whole organization is structured is a separate and larger question. One workflow is enough to expose the pattern.

One workflow, redesigned from the outcome back

This is an illustrative construction, not a customer case. Take a billing dispute handled by a service team.

Today’s task list

A customer calls about a charge they do not recognize. The task list looks like this:

  • Log the case in the service system.
  • Look up the account in the billing system.
  • Re-enter the charge details into a dispute form.
  • Send the form to the billing team.
  • Wait for the billing team to review and respond.
  • Request supervisor approval for any credit above a set amount.
  • Apply the credit manually.
  • Email the customer a confirmation.
  • Pull a sample of closed cases for quality review.

Score it for automation and the top candidates are obvious: the re-keying, the form routing, the confirmation email. Automate those and the spreadsheet declares victory.

Starting from the other end

Beneficiary and outcome. The customer wants to know, in one contact, whether the charge is right, and to have it corrected if it is not. Written as a sentence: the customer leaves the contact with a correct answer, a correct balance and an explanation they believe.

Required contributions. Working back from that sentence:

  • An accurate picture of the account and the charge, in one place.
  • A decision: is this charge an error, or legitimate?
  • If an error, a correction applied within policy.
  • An explanation the customer can follow.
  • Assurance that the decision matched policy and the customer’s real situation.

Allocation.

Required contribution Allocated to Why
Assemble the account and charge picture Agent Information gathering from known systems
Decide: error or legitimate Person, supported by the agent’s summary Ambiguous cases need judgment and accountability
Apply a correction within policy Agent for clear cases, person for exceptions Settled rules versus exceptions
Explain to the customer Both The agent drafts, the person adapts to the customer
Assure the decision Both Checks built into the flow instead of sampled afterward

What is left over. Compare this with the original list. The re-keying, the form, the handoff to billing and the wait have no contribution to supply. They existed because the service team could not see billing data. The supervisor approval for small credits existed to guard against inconsistent decisions, which clear policy bounds and built-in checks now handle. The after-the-fact sample review existed because assurance was not part of the flow.

Those tasks did not get automated. They disappeared.

What remains is a small number of contributions, and one of them is a hard one: deciding well when the charge is ambiguous. That is where capability becomes a real question.

Where capability enters, and why we diagnose first

Only now do you know what the work requires from people. In our example it is operational judgment on ambiguous disputes, not recall of a product manual. That changes everything about what you might build.

But even here, resist the reflex to build a course. A U.S. Office of Personnel Management resource says to define the performance problem, identify the critical behaviors and why they are not happening, and to choose training only if it is the best response. Other responses include communicating expectations, providing a supportive work environment or checking job fit.

The International Finance Corporation’s guide makes a similar point: individual performance depends on clear expectations, feedback, resources, skills, knowledge and motivation. Only some of those are learning questions.

In plain terms: ask whether the gap sits in the person or in the environment. If agents decide badly because the agent summary hides the key fact, no course fixes that. Fix the summary. If they decide badly because they have never practiced the borderline cases, practice is the answer.

Our diagnosis-first playbook for skill gaps and our piece on performance support versus training go deeper on that check.

Where capability enters, and why we diagnose first

What counts as evidence

By evidence we mean your organization’s own source material, the stuff that grounds a design decision. The OPM planning resource points to strategic plans, HR metrics, job descriptions, surveys and performance appraisals as inputs. For operations work, we add what sits closest to the work:

  • Strategy. What the business is trying to change, and for whom.
  • Frontline friction. Where people work around the system, escalate, or re-do things.
  • Quality findings. Where outcomes miss the standard, and the reasons recorded.
  • Expert knowledge. How your best people actually make the call.

Not every gap deserves attention first. One utility’s knowledge toolkit recommends prioritizing knowledge gaps by risk and criticality to safety, operational performance, compliance and customer needs. That is a sound filter for deciding which contributions to protect first.

Building for judgment, not recall

When the diagnosis says people need to develop judgment, the design should resemble the decision. A university chapter on competency-based training advises designing from job analysis and the knowledge, skills and abilities required for effective performance. It also notes that scenarios are richer complements to written tests.

For our billing example, that means practicing real borderline disputes drawn from your own quality findings, with the facts a person would see and the trade-offs they would weigh. It does not mean a slide deck on dispute policy. The policy is a constraint on the decision, not the decision.

Perspective

We hear the same request every week: help us build training faster. We understand it. The pressure is real and the tools are tempting.

But speed at the wrong starting point only produces more of the wrong thing. If the work itself carries handoffs nobody needs, a faster course teaches people to operate a process that should be gone. If the gap sits in the environment, no amount of training closes it.

That is why Cognistry starts earlier. We diagnose what capability the work requires, decide whether learning is the right enablement play at all, and ground each design decision in your own evidence: strategy, frontline friction, quality findings and the expertise of your best people. Courses, practice simulations and decision environments come after, built to develop operational judgment and measured back to the business.

Redesign from the outcome back. Ask what the system needs contributed. You will find work that deserves to be automated, work that needs people who decide well, and work that simply goes away.

Brian

How Cognistry fits

Once you know which contributions need people and what judgment they require, you need a way to ground the answer in evidence, structure the response and show it worked. That is what the Cognistry platform is for. Signal maps the capability signals in your own source material. Forge structures the learning architecture. Sim builds decision practice environments for operational judgment.

Cognistry

If you are mapping a workflow now, bring it to a conversation. We will start with the outcome, not the course. See the platform overview to get started.

Proving it worked

A redesign needs a credible line from the change to the business result. The U.S. Office of Personnel Management says metrics, benchmarks and targets are needed to judge whether workforce plans meet mission needs.

BCG’s approach to measuring upskilling has three steps: identify the target business impact, define time-bound metrics and milestones, and determine whether the impact was achieved. While business impact takes time to appear, it suggests using interim signals.

For the billing workflow, define the outcome measures before you redesign: how often a customer’s dispute is settled in the first contact, how often a settled case is reopened, and how decisions line up against policy. Tie them to the outcome you wrote in step one. Then watch whether behavior in the work changes, not only whether people completed an activity. McKinsey’s guidance on capability building for transformations similarly advises tying capability to business value drivers and measuring observed behavior change and operational impact.

Proving it worked

We cover a practical window in our guide to measuring training impact in 60 to 90 days, and on the speed side, the time to competence metric. Teams running operations workflows can also look at our page for operations teams.

A short checklist before you automate or build

  • Can you state the beneficiary and outcome in one sentence?
  • Have you listed required contributions without naming today’s tasks?
  • For each contribution, have you decided: person, agent or both?
  • Which current tasks supply no required contribution?
  • For the people contributions, is the gap in the person or in the environment?
  • Do you have your own evidence for the answer, not just opinion?
  • Have you defined the business measures before you launch?

FAQ

What does “start with what the system needs contributed” mean?

It means designing work from the result backward. You define what the beneficiary needs, work out what must be supplied for that to happen, and only then decide who or what supplies it. A task is one way of delivering a contribution, not the unit you begin with.

What are the three parts of a system?

The simplest breakdown is input, process and output. For work, that is what enters (requests, information), what transforms it (people, tools, steps) and what the beneficiary receives. Work system research goes further and lists nine elements, including customers, participants, information, technologies and environment.

Is a task-automation spreadsheet useless?

No. It is a good way to see what people do today and to find quick efficiency gains. Its limit is that it starts from existing tasks. Use it after the outcome-back pass, to decide how to supply the contributions you kept.

When is training the right answer?

When the diagnosis shows people lack the knowledge or judgment the work requires, and the environment already supports them. Federal guidance on training needs advises selecting training only if it is the best response to the defined problem. Otherwise, other enablement plays such as clearer expectations or a better work environment come first.

Sources