AI Strategy5 November 20268 min read

The AI Productivity Dividend — Part 3 of 6

Hire the outcome, not the attendance pattern

Most job adverts describe when someone must be present before they describe what the person must achieve. Part 3 of The AI Productivity Dividend shows how to write and fill a role from its outcomes, test every attendance requirement, and widen access to scarce skills without lowering the bar.

#Operating Model#Leadership#AI Roles#Work Design
A single stream of warm light enters a brushed-metal manifold and divides into five channels of different widths, one fitted with a small amber valve.

When Hantam Service Cooperative, a fictional shared-services cooperative in the Northern Cape, last advertised for a cloud cost analyst, the advert's first line was "Monday to Friday, 08:00 to 17:00, office-based in Calvinia". The skills came in the third paragraph. The shortlist was short, and the person appointed left within eighteen months.

This year the cooperative needed more cloud cost capacity again. Eighteen new member organisations had joined, existing members were moving more workloads to the cloud, and the existing analyst had moved to a three-day role in the pilot described in Part 1. The work that needed doing was measured at roughly three days a week. The old advert would have offered five, or nothing.

The question this article answers is how to hire for the work rather than the attendance pattern: how to write a role from its outcomes, how to test which attendance requirements are real, and how to do that without lowering the bar or creating a second-class tier of employees.

Attendance is a proxy, and a weak one

Fixed attendance is a convenient stand-in for things that are harder to specify: that someone will be reachable when needed, that they will absorb context from colleagues, that they are committed. Each of those is a real need. None of them is the same thing as being present five days a week.

The trouble with the proxy is that it filters candidates on availability before it filters them on capability. It excludes people who can do the work extremely well but cannot, or do not want to, sell five full days every week. Part 5 looks at who those people are. This article is about the mechanics of not filtering them out by accident.

The fix is the same one that runs through CloudNala's work on roles: describe the outcome and the evidence, not the activity. AI roles decoded: read the verbs, not the title applied that idea to job titles. Here it is applied to time.

Write the role from the outcome

Hantam wrote the new role profile in five layers, in this order.

Write the role from the outcomeLAYERHANTAM'S SECOND ANALYST1 Accepted outcomesAllocations for 18 new members2 Service levelsReports by working day 53 Availability windowsClose days; two-hour overlap4 Decision rightsSigns allocations; recommends buys5 Evidence of performanceAccuracy, timeliness, usefulnessThen, and only then: three days a week, days agreed around the close
The role profile is written in five layers, starting with what must be delivered and ending with how performance will be judged. Only then is the working pattern stated. Availability windows say exactly when someone must be reachable and why, which is a much narrower requirement than five days of presence.
  1. Accepted outcomes. What the role must deliver, stated so that someone can tell whether it was delivered. For the second analyst: cost allocations for the 18 member organisations that joined this year, quarterly rightsizing recommendations with evidence, and the cooperative's commitment-purchase analysis twice a year. "Accepted" is deliberate, in the same sense as cost per accepted outcome: work counts when its recipient accepts it, not when it is produced.
  2. Service levels. How quickly and how reliably: reports by the fifth working day, member cost questions answered within two working days.
  3. Availability windows. The specific times when someone must be reachable, and why. For this role: the close at the start of each month, and a two-hour overlap on working days for collaboration with the platform team.
  4. Decision rights. What the role decides alone, what it recommends, and what it escalates. The analyst signs off allocations for her members and recommends commitments; the finance director approves purchases.
  5. Evidence of performance. How the role will be assessed: accuracy of allocations, timeliness, the quality of recommendations as judged by the people who act on them.

Only after those five layers did the profile state the working pattern: three days a week, with the days agreed at appointment around the close and the overlap window.

Test every attendance requirement

Every requirement about when or where someone works should survive a short test before it goes into an advert.

Test every attendance requirementAttendancerequirementDoes an acceptedoutcome need it?yesCan a rota, handoveror record meet it?noStill essentialto the job?noRemove ityesMove it to thatmechanismyesKeep it andwrite down whyno: remove or redesign
Each requirement about when or where someone works is tested in turn. If no accepted outcome needs it, it goes. If a rota, a handover or a written record can meet the need, it moves there. What survives is written down with its reason. This is a design test, not legal advice on inherent requirements.
  • Does an accepted outcome require it? The month-end close does require presence on particular days. "Available for ad hoc meetings" does not trace to any outcome.
  • Could a rota, a handover or a written record meet the need instead? Same-day acknowledgement of cost anomalies is a real need, but Hantam's rota meets it. It does not need to sit in this role.
  • Is it still essential to the job? If, after the first two questions, the requirement still stands, write it down with its reason.

This test has a legal dimension in South Africa. Section 6 of the Employment Equity Act prohibits unfair discrimination, directly or indirectly, in any employment policy or practice, on listed grounds that include family responsibility, age and disability, and it covers applicants too, as outlined in Hire on evidence. Section 6(2)(b) says it is not unfair discrimination to distinguish, exclude or prefer a person on the basis of an inherent requirement of a job. Whether a particular attendance rule could amount to indirect discrimination in a given case is a question for legal advisers, not for this article. The design point is simpler: an attendance requirement that cannot be traced to an outcome is both a weak filter and an unnecessary risk. This is general information, not legal advice.

Write an advert that tells the truth

A reduced-time advert fails when it is vague. "Flexible hours negotiable" tells a candidate almost nothing and invites the suspicion that flexibility will be granted grudgingly. Hantam's advert said:

  • what the role must achieve, in the outcome language above;
  • the working pattern: regular days of Wednesday to Friday; in close week, the first three working days of the month instead, dropping other days, as agreed in writing;
  • the pay, stated both as the actual salary and as the full-time-equivalent rate, so that candidates could see the hourly rate matched the full-time analyst's;
  • what was not expected: no availability on non-working days, no attendance at meetings outside the overlap window;
  • benefits and progression, including that the role sits on the same career ladder as full-time analysts.

The advert did not ask candidates why they wanted three days, and the interviews did not either. Reasons can involve health, family or other matters that are none of the employer's business, and some of that information, such as health, is special personal information under section 26 of the Protection of Personal Information Act. The employer needs to know whether the candidate can deliver the outcomes in the pattern offered. That is all.

Select on the outcomes, including the quiet ones

Hantam assessed candidates with work samples traced to the role's outcomes, using the approach in Hire on evidence: interviewing for AI-era roles: a stated rule on AI use, anchored scoring and a decision record. Candidates were allowed to use an AI assistant in the cost-allocation exercise, because the role uses one, and were scored on whether their figures were right and their reasoning visible.

One exercise was specific to reduced-time work. Candidates received a half-finished monthly allocation with a short trail of messages and were asked to write the handover note they would leave for a colleague covering the next two days. A reduced-time role depends on the quality of what is written down between working days, so this is a core skill, not a soft one. Part 4 explains why.

Pay, progression and influence

A reduced-time role that pays fairly but offers no progression or influence is still a lesser job. Four commitments prevent that.

  1. The same hourly rate as a comparable full-time role, and pay that moves on the same cycle. Where section 198C of the Labour Relations Act applies, it requires part-time employees to be treated, taking hours into account, on the whole not less favourably than comparable full-time employees unless there is a justifiable reason, and to have the same access to opportunities to apply for vacancies. Fair employers apply that principle whether or not the section covers the role. Naledi's arrangement in Part 1, 75% of salary for 60% of the days, is a time-limited share of a verified gain agreed for her pilot, not a different rate for the role; the second analyst's hourly rate is the standard one.
  2. Promotion on outcomes, not visibility. Hantam's analyst ladder now describes each level by the scope and difficulty of outcomes owned, not by availability or attendance.
  3. Access to training on the same basis as full-time colleagues, scheduled on working days.
  4. A real voice. Reduced-time staff are invited to the decisions that affect their work, at times they work, and their written input counts as much as a spoken contribution in a meeting.

Measure AI capability through work products makes a related argument about capability: judge what people produce, not how they appear. The same principle protects reduced-time staff from being judged on presence.

What happened at Hantam

The figures here are illustrative. The outcome-first advert drew more applications than the previous five-day advert, and a wider range of experience, including candidates with deeper cloud cost expertise than the old shortlist. The cooperative did not record why any candidate wanted three days, and the appointment was made on the work samples. Six months on, the second analyst's allocations are as accurate as the full-time benchmark, and her handover notes have become the template for the team.

The larger effect was on the cooperative's managers. Writing the role from its outcomes made them notice how many five-day roles had never been described that way. Two other posts are now being rewritten the same way, whatever their final working pattern.

Checklist for an outcome-first role

  • The role profile starts with accepted outcomes, service levels, availability windows, decision rights and evidence of performance, in that order.
  • Every attendance requirement traces to an outcome or is removed, and those that remain are written down with their reason.
  • The advert states the actual pay, the full-time-equivalent rate, the working pattern and what is not expected.
  • Candidates are not asked why they want reduced hours.
  • Selection uses work samples, including one that tests written handover.
  • Pay rate, progression, training and influence are on the same basis as full-time roles, and HR and employment-law advisers have reviewed the terms.

Key points

  • Fixed attendance is a proxy for reachability, context and commitment. Design for those needs directly.
  • Write the role in five layers: outcomes, service levels, availability windows, decision rights and evidence. State the working pattern last.
  • Test every attendance requirement against an outcome, an alternative mechanism and the inherent requirements of the job.
  • Tell the truth in the advert: actual pay, equivalent rate, pattern and what is not expected. Do not ask for reasons.
  • Equal hourly pay is not enough. Progression, training and influence must not depend on presence.

Next: Part 4, The hidden architecture of part-time knowledge work. All parts are on the series page.

How CloudNala can help

CloudNala helps technology and operations teams rewrite roles around outcomes when AI changes the work. We turn a job description into an outcome-first role profile, run the attendance requirement test with your managers, design work samples that test the skills a reduced-time role really depends on, and check the role against your career ladder, alongside your HR and employment-law advisers.


Work with CloudNala

CloudNala helps organisations move from technology ambition to practical execution across cloud, AI, data, platform engineering and digital services.

Whether you are exploring AI, modernising your cloud environment, building a public-sector digital service, or turning an idea into a working MVP, we can help you shape the roadmap and deliver the next step.

Talk to us about AI and work design or write to us at consult@cloudnala.co.za