Read this for your industry
Optional. Choose your industry and this piece opens with a situation from your world and closes with a first move for you. Your choice carries across the guides and analysis on this site.
You lead advice across a licensee
Australia had almost 29,000 financial advisers at the start of 2019 and 15,127 in March 2026, according to the Financial Advice Association Australia. The same release cites Investment Trends research that 15.9 million Australians have unmet advice needs, and notes that only 464 of today's advisers entered the profession in 2025. Over the coming decades, around $3.5 trillion is expected to pass from one generation to the next.
If you lead advice for a licensee, those numbers become one question for you: how can each adviser look after more people while the quality of advice holds? Agents can take on much of the preparation that fills an adviser's week, from fact-find summaries to research and first drafts of a Statement of Advice. The way you brief and check them will shape the work of every adviser in the licensee, and the experience of thousands of clients.
What this means for you
- Run agents as a shared paraplanning service for the whole licensee, with one written standard for each piece of work, so every adviser starts from the same quality of preparation.
- Use your compliance file reviews as the quality bar. When reviewers find the same gap across several advisers, it becomes a rule the agents follow.
- Keep the adviser as the person who gives the advice and answers for it. The agents prepare the work, and the adviser decides what the client hears.
- Measure what clients feel: the time from first meeting to advice and the number of clients each adviser can look after, side by side with file quality.
You run a territory of a dozen stores
In September 2025 the Federal Court found that Woolworths and Coles had underpaid salaried store managers and had not kept the records the law requires, including records of overtime. Russell Kennedy puts the number of managers covered at more than 27,000, and Woolworths estimated a pre-tax liability of up to $750 million, as SBS reported. Store teams are also facing more abuse. In the SDA's 2025 member survey, 25% had faced physical violence in the past year and 60% wanted more follow-up from their managers.
As an area manager, you stand between those pressures. You approve rosters, you see the hours your store managers really work, and you decide how incidents are followed up. Agents can draft rosters, keep the records and track every incident through to its follow-up. The standard you set for them decides how much safer and fairer your stores become.
What this means for you
- Brief rostering agents to work to the award and to keep a written record of every change, including overtime for salaried managers. The court case turned on records that were never kept.
- Give incident follow-up to an agent that drafts the check-in and tracks it until a manager has had the conversation. The conversation itself stays with a person.
- Ask agents to compile patterns across your stores, such as which stores keep running over hours or which keep reporting the same offenders, so your visits go where they are needed.
- Use the time saved to put store managers back on the floor with their teams.
You lead operations across a group of aged care homes or clinics
The new Aged Care Act and the strengthened Quality Standards took effect on 1 November 2025, including a dedicated standard for clinical care. Homes are working to a sector average of 215 care minutes per resident per day, including 44 minutes from a registered nurse, and from the 2025-26 financial report every residential provider must submit a Care Minutes Performance Statement audited by a registered company auditor.
In general practice, the RACGP reports that GPs' use of AI scribes doubled in a year, and its guidance says every scribe output must be carefully checked because these tools can make errors.
If you lead operations across several sites, you carry both pressures at once: more evidence to report, and staff who already spend too much of each shift on documentation. Agents can draft care-plan reviews, compile care-minute evidence from rosters and prepare referral letters. How you brief and check them decides whether nurses and carers get that time back with residents and patients.
What this means for you
- Start with reporting work, where agents can assemble care-minute and registered nurse coverage evidence from rosters, and your finance lead checks it before it goes to the auditor.
- Keep clinical documentation under clinical review. Agents draft, and the nurse or doctor who answers for the record approves it.
- Use one standard across every site. When audit findings repeat at several homes or clinics, they become rules for the agents everywhere.
- Write consent and privacy into every brief, as the RACGP's practice standards ask of general practices using AI.
You lead engineering or sourcing across several teams and suppliers
Google's 2025 DORA report, drawing on nearly 5,000 technology professionals, found that 90% now use AI at work and that AI adoption is linked to higher delivery throughput. It also found a continuing negative relationship with delivery stability, and concluded that AI amplifies whatever a team already has, good or bad.
In supply chains, Ai Group found 47% of Australian industrials reporting active disruptions in August 2025, and 44% of manufacturers planning to raise their investment in supply chain resilience in 2026. Entities with annual revenue of at least $100 million must also report each year on how they assess and address modern slavery risks in their supply chains.
If you lead engineering, agents will soon write more code than your reviewers can read, unless you change how work is checked. If you lead sourcing, agents can watch hundreds of suppliers for risk and compare quotes overnight. In both cases, the standards you set decide how much of that speed reaches your customers.
What this means for you
- Make automated tests, the definition of done and code review the quality bar for every agent. DORA's finding on stability shows what happens when output grows faster than the checks.
- Give each agent one role. The agent that writes the code should never be the one that approves it.
- In sourcing, let agents monitor suppliers and draft due diligence, and keep decisions about suppliers with the people who own those relationships.
- Turn every repeat defect into a corrective action in the agent's instructions, the same way you would with a supplier.
You run operations for a growing solar and battery installer
More than 457,000 batteries were installed in the first 12 months of the Cheaper Home Batteries Program, and installations have since passed 500,000. The number of accredited battery installers doubled to 8,846, and Jobs and Skills Australia estimates the country will need about 32,000 more electricians by 2030.
Quality has struggled to keep pace. Between July 2025 and April 2026, the Clean Energy Regulator inspected 1,278 battery installations and found 60.8% substandard and 1.2% unsafe. Most of the problems were missing or incorrect labels, alongside more serious faults such as exposed wiring.
If you run operations for an installer, you are hiring fast, training new crews and answering customers who have waited months. Agents can prepare quotes, check designs against the brief and review job photos before a crew leaves site. A well-written brief can catch the labelling and paperwork gaps that inspectors find most often.
What this means for you
- Brief a review agent with your compliance checklist, including labels, backup circuits and photo requirements, and have it check every job's photos the same day.
- Keep every electrical sign-off with the licensed person who answers for it.
- Turn every defect that an inspector or your own reviewer finds into a rule in the job pack, so the next crew does not repeat it.
- Use agents for quotes and customer updates, so your designers and electricians spend more of their time on design and on site.
The first month in a new territory
If you have ever taken on a territory as an area manager, you will remember the first month. You have ten or twelve stores, each with its own manager, and no two of them run quite the same way. One manager keeps a tight Saturday roster. Another has a stockroom nobody has sorted since the last stocktake. The stores are trading and the people know their jobs. Your problem is that you can only stand in one store at a time.
The area managers who do well learn to run the territory through other people. They write the standard once and take it to every store. They use each visit to check the few things that matter, and they pay attention when the same problem appears in three stores, because that usually points back to the process. They also shape the culture of every store they visit, often more than they realise, because store managers copy what their area manager pays attention to.
Running a team of AI agents asks for the same habits. You cannot watch every agent at work, so you rely on clear roles, written standards and well-chosen checks. The one difference is that this team forgets everything overnight.
Where the name comes from
Microsoft’s 2025 Work Trend Index gave this a name. It described the agent boss as someone who “builds, delegates to, and manages agents to amplify their impact”, and predicted that every worker would need to think like the head of a small team of specialist agents.
I like the name because the second word carries the weight. Being an agent boss is mostly management.
At the start of 2026 I set up a coordinator, a researcher, a writer, a reviewer and a couple of specialists to help with my own writing and research. I had a working system within a month. Within three months I had noticed that almost every problem I hit was one I had already seen in teams of people: unclear roles, vague briefs, the same mistake made twice, and a capable person doing someone else’s job.
What an agent boss does
The job has a lot in common with an area manager’s.
- Gives each agent one role. When my researcher also wrote the final document, both jobs got worse.
- Writes the brief. What is needed, what good looks like, the format and the limits.
- Sets the quality bar before the task starts, the way an area manager agrees the store standard before the visit, so the review is a check against something both sides know.
- Notices repeat mistakes and writes them down as rules. A mistake that keeps coming back is like the problem that appears in three stores: the instructions need fixing.
- Decides where a person must check the output, and who answers for the result.
Almost all of that is ordinary management. It is the list you would write if you had just inherited five capable graduates who start each morning with no memory of the day before.
The four stages
Looking back, my setup went through four stages. I have since seen the same pattern in organisations I talk to.
1. Operator. One person uses one or two agents for single tasks. Corrections happen in the chat and are lost when it closes, so you pay for each lesson more than once. This is the area manager who fixes every problem personally on each visit, so each store is only as good as the last visit.
2. Manager. Corrections become written rules that the agents load every time. New people inherit the rules and skip the rediscovery. This is the territory where every store manager works from the same written standard. It is where I saw the first real gains, and where I spent most of my first three months.
3. Architect. You redesign the work around what agents do well and put people where judgement matters. Governance gets written down: who can set up an agent, what it can decide alone and what triggers a review. In a territory, this is the area manager deciding which tasks move to a central team and which decisions stay with the store manager.
4. Strategist. Agent capability becomes part of the operating model, and hiring, training and performance measures change to match. In retail, this is the point where the region changes how it recruits and develops its store and area managers.
MIT CISR’s Enterprise AI Maturity Model, based on a survey of 721 companies, describes four stages of enterprise AI maturity and found that financial performance improved at each one. I think of the agent boss stages as the same climb, seen from one desk.
Where most people get stuck
Most people stop at stage one, and I understand why. Using AI to do a task faster gives you a result you can see straight away. Writing a brief and a set of rules feels slower in the week you do it, much like an area manager writing up the standard when they could fix the display themselves in ten minutes. The store manager who watches them fix it learns that the area manager will always step in.
Moving to stage two means asking a different question. At stage one you ask how AI can help you with a task. At stage two you ask how to brief an agent so it can own that kind of task every time. The second question is a management habit, and most organisations have not trained people for it.
BCG’s AI at Work 2025 survey points the same way across thousands of workers. About half of companies are moving beyond productivity gains to redesign workflows, and employees in those companies save more time and spend more of it on strategic tasks. Regular use was much higher where people had at least five hours of training, yet only about a third said they had been properly trained.
Training people to brief and review agents is the cheapest way I know to move a team from stage one to stage two.
Why area managers matter most
If you are planning how agents will work in a business with a large frontline, I would start with the area and territory managers. Store managers are close to the work and senior leaders set the direction. The area manager sees enough stores to spot a pattern and has the authority to change the standard. Their habits spread as well, because when an area manager briefs well and checks the right things, a dozen store managers see it every week and start doing the same.
The same holds in most frontline businesses. The practice principal in a wealth business, the operations lead across a group of clinics and the manager who dispatches installation crews all sit at that point where decisions reach the most people.
What this series covers
Each piece takes one problem I hit, what it cost and the fix that held.
- The correction log: why a correction made in conversation fades, and how to make it last.
- The delegation trap: why capable people and capable agents keep doing tasks they should hand on.
- The confidence trap: what to do when an agent sounds certain and is wrong.
- Culture comes before agents: what agents reveal about the habits a team never wrote down.
Where to start this week
Go back to that first month in a new territory. The area managers who settle in fastest are the ones who start writing the standard early.
Pick one task you do every week that an agent could own, and write the brief as you would for a new starter. The third time you correct the same thing, write the correction down as a rule. That is your first step from operator to manager.
Find the step where clients wait longest, which in many practices is preparing the Statement of Advice. Write the licensee standard for that step once, run it with a small group of advisers for a quarter, and compare their file review results with the rest of the network.
Start with rosters and hours records for your territory. Write one brief with your payroll team, check the first month of agent output against the award, and make sure every roster change has a record behind it.
Pick the report your sites find most painful each quarter, which is often care minutes and registered nurse coverage. Write the brief once, run it across two sites with a finance check, and measure the hours of nursing time it gives back.
Pick one product team or one supplier category. Write the brief and a worked example of good output, track change failure rate or supplier defects for a quarter, and widen the rollout once the numbers hold.
Take the last ten jobs your team completed and run their photos and paperwork past the checklist. Every gap you find becomes a line in the brief, and from next week the review agent checks each job before the crew drives away.
Questions readers ask
Does the agent boss idea work outside office jobs?
Yes. The habits come from frontline management: clear roles, written standards, good handovers and checks before anything reaches a customer. You can read this guide for wealth management, retail, healthcare, product engineering and manufacturing, or new energy and trade services by choosing your industry at the top of the page.
Where does the term agent boss come from?
Microsoft used it in its 2025 Work Trend Index for someone who builds, delegates to and manages agents. I use it the same way, with the weight on the managing.
Do you need to be technical to be an agent boss?
No. The hard parts are writing a clear brief, deciding what good looks like and noticing when the same mistake comes back. Most experienced managers already have those skills, and they transfer well.
How long does it take an organisation to get good at this?
In my experience the first useful agent takes a few weeks, and the habits that make agents reliable across a team take months. MIT CISR describes enterprise AI maturity as four stages, with financial performance improving at each one.
Sources
- 2025: The year the Frontier Firm is born (Work Trend Index), Microsoft WorkLab, April 2025
- Building Enterprise AI Maturity, MIT CISR, December 2024
- AI at Work 2025: Momentum Builds, but Gaps Remain, BCG, June 2025
The views here are my own, not those of my employer or its clients.
More in The agent boss
The correction log: why AI agents repeat the same mistakes
A correction made in conversation stays in that conversation, so AI agents repeat the same mistakes. Writing corrections down as rules is what makes them last.
The agent bossThe delegation trap: why capable people and agents keep doing the work
Strong performers struggle to delegate because doing a task themselves is faster this week. AI agents fall into the same habit. How to spot it and how to break it.
The agent bossThe confidence trap: when an AI agent is certain and wrong
AI agents sound equally sure whether they know or are guessing, and people tend to trust that fluency. Two standing rules that make agents say what they don't know.
The agent bossCulture comes before agents: what AI exposes about how you work
AI agents do exactly what they are told, so they find every habit a team never wrote down. Why culture comes first with agents, and how to make it explicit.