Telstra: eight years of transformation, and the operating model still can't keep up

Telstra has run three named transformation strategies in eight years and can show profits, dividends and an AI-trained workforce for it. In July 2026, a defect its own technicians logged in 2022 took the national network down for most of a day. This episode reads the gap between those two facts.

In 2018, Telstra CEO Andy Penn stood in front of investors and announced T22, a four-year transformation strategy he described as "a strategy of necessity." Telstra's consumer business was bleeding. NBN was gutting its fixed-line margins. Mobile competition was intensifying. The company needed to simplify, digitise, and cut costs or face a slow decline into irrelevance.

T22 was ambitious. Reduce 1,800 consumer plans to 20. Digitise customer interactions. Cut around a quarter of the workforce. Hire 1,500 digital specialists. Build a technology platform that could carry the company into a different kind of future.

By 2022, Telstra declared T22 complete. The plan simplification was real. The workforce reduction was real. The digital hires were real. On its own terms, the strategy delivered.

Then, nine months before T22 formally closed, Penn announced T25, "a strategy for growth." New targets: A$500 million in cost savings, 95% 5G population coverage by FY25, and a pivot from defensive simplification to offensive expansion.

Penn stepped down in September 2022. Vicki Brady, his successor, inherited T25 and the organisational machinery that came with it.

The pattern emerges

What happened next is worth watching carefully, because it is a pattern that shows up in large organisations more often than anyone admits.

In May 2024, Telstra announced 2,800 job cuts, roughly 9% of its 31,000-strong workforce. The A$500 million cost reduction target from T25 was quietly revised downward to A$350 million. The 95% 5G coverage target that had been central to the T25 announcement stopped being restated in investor materials. When the next strategy arrived, it carried a different number altogether: 91%, by FY30. Nobody stood up and said: we set a target, we did not hit it, here is why. The target was not missed in public. It was replaced.

In January 2025, Telstra signed a A$700 million, seven-year joint venture with Accenture, which the company would later describe as compressing "a five-year AI roadmap into two years." The partnership was framed as transformational. The price tag suggested Telstra did not believe it could build AI capability internally, despite having spent six years hiring digital talent.

Then, in May 2025, came the third strategy in seven years. Connected Future 30, or CF30, replaced T25. New branding. New targets. New language about AI-native operations and customer experience. The structural question nobody asked publicly: if T22 was complete and T25 was on track, why did the strategy need replacing?

The numbers look good

And this is where the Telstra story becomes instructive, because on the surface, the numbers tell a success story.

FY25 results, reported in August 2025, showed net profit of A$2.34 billion, up 31%. The company reported hundreds of AI use cases in deployment; by February 2026 the count stood at 380. It had cut 3,208 roles across the year. The profit line was moving in the right direction.

By H1 FY26, reported in February 2026, profit reached A$1.2 billion, up 8.1%. Calendar 2025 alone accounted for another 2,356 roles. The workforce was heading toward 29,000, down from around 32,000 when T22 began.

The full year confirmed the trajectory. FY26 results, released on 13 August 2026, showed net profit up 2.7% to A$2.4 billion and earnings per share up 5.3%. The final dividend lifted the full-year payout 10.5% to 21 cents. Telstra completed a A$1.25 billion share buyback during the year and announced another of up to A$1 billion. Revenue slipped slightly, and the shares fell on the day, but the headline story held: profits up, dividend up, capital returned.

If you looked only at the financial statements, you would see a company executing well. Costs were coming down. Profits were going up. AI was being deployed at scale.

But the financials measure the output of decisions. They do not measure the quality of the system making those decisions. And inside Telstra, something was not working.

The confession from within

In April 2026, Telstra International CEO Roary Stasko proposed cutting 128 internal roles and 35 contractor positions. The reason he gave was revealing. Stasko cited "complex and inefficient processes" including "manual workarounds and operational backlogs," in internal messages reported by Information Age.

Read that again. In April 2026, eight years into continuous transformation, a division CEO was naming manual workarounds and operational backlogs as the reason for restructuring. Not legacy technology. Not market shifts. Process complexity. The kind of friction that transformation programmes are specifically designed to eliminate.

This is not a failure to transform. It is something more uncomfortable. It is evidence that the transformation addressed the visible layers (plans, headcount, digital tools) without resolving the structural layer underneath. The plumbing changed. The flow did not.

The outage

On 8 July 2026, routine maintenance to replace faulty backup power at a Melbourne data centre required restarting one of the servers that keep time across Telstra's network. Because of a known software defect, the server came back up believing the year was 2006. The wrong date spread through the network, authentication certificates failed, and the fault cascaded into a nationwide outage. Service was disrupted for much of the day, and Telstra later wrote to 8.8 million affected consumer and small business customers. Over the course of the outage, 58,835 Triple Zero calls connected successfully and 604 failed with an error. Telstra completed 604 welfare checks on affected callers, and police completed all 144 in-person follow-ups with no adverse outcomes reported. Victoria's entire regional rail network stopped.

The defect was not a surprise. Information Age reported that the vendor had issued advisories about the bug years earlier, that Telstra technicians logged it internally in September 2022 and a risk assessment concluded it was not relevant to the network, and that the fix was declined again in January 2026. Vicki Brady told a Senate hearing in July that the outage was caused by a missed software update, alongside an intentional design change to the time-keeping hardware that had not been properly documented. The ACMA opened an investigation into Telstra's compliance with its emergency-call obligations.

Eight years into continuous transformation, the country's largest telco was taken down by a maintenance task whose risk was already logged inside the building. That is not a technology story. It is a signal-flow story, dated 8 July 2026.

The workforce paradox

In June 2026, Dr Mark van Rijmenam published a WAVE assessment of Telstra (Watch, Adapt, Verify, Empower, his framework for AI-era workforce readiness) that gave the company an overall score of 8.8 out of 16. His summary was precise: "Telstra has built the skills base of an AI-fluent company and the decision architecture of a 20th-century carrier."

The detail beneath that headline is where the real story sits. Telstra had trained more than 22,000 employees in AI. It had deployed 18,000 Microsoft Copilot licences with 80% weekly active usage. Its workforce was, by any reasonable measure, ahead of the curve.

But trained in what, exactly? And to what effect? Van Rijmenam's findings cut to the core. "Empower without Adapt is a coiled spring," he wrote. Trained staff without authority, in his words, "become flight risk."

Telstra had invested heavily in building capability. It had not invested correspondingly in changing the authority structures that determine whether capability can be applied. People knew how to use the tools. They did not have permission to change the processes those tools were meant to improve.

When asked directly in February 2026 whether AI had replaced any specific roles, CEO Vicki Brady could not point to one. The technology was "making the business more efficient," she said, but she could not connect that efficiency to a concrete structural change. Efficiency without structural change is optimisation. It is making the existing system run faster, not building a different system.

The naming problem

Step back and look at the sequence: T22, then T25, then Connected Future 30. Three strategy brands in seven years. Each announced with conviction. Each carrying new targets, new language, new framing.

The naming is not cosmetic. It reveals something about how Telstra relates to its own transformation. Each rebrand creates a clean narrative break. The new strategy does not have to account for the unfinished business of the old one. T25's 5G coverage targets can disappear without a public reckoning because T25 itself has been superseded. The accountability resets with the brand.

Penn, speaking to McKinsey in September 2022 as T22 wrapped, offered a line that reads differently now: "If you allow ambiguity to filter into what you're trying to do, suddenly you lose control of the agenda."

The irony is that Telstra's transformation has been anything but ambiguous in its stated intent. The strategies have been clearly articulated, professionally communicated, backed by significant investment. The ambiguity is structural. It lives in the gap between what the strategy says the organisation is becoming and what the operating model actually permits.

The read

Telstra is not a failing company. It is a profitable, operationally capable telecommunications provider that has successfully simplified its consumer business and invested seriously in AI capability. That makes it a more useful case study than a dramatic collapse, because the pattern at work here is the one most organisations actually face: the transformation looks right from the outside and feels wrong from the inside.

Here is what the four coherence questions reveal.

Agency: can the people closest to the work act on what they can see?

Telstra trained 22,000 people to use AI tools. It gave 18,000 of them Copilot licences with 80% weekly usage. Those people can see opportunities for improvement. The question is whether they can act on them.

Van Rijmenam's assessment says no. The authority structures have not changed at the same rate as the capability. You have a workforce that is increasingly skilled and increasingly constrained. They can see what needs to change. They cannot change it without navigating the same "complex and inefficient processes" that Stasko named in his restructuring proposal.

This is a specific and dangerous failure mode. When people are trained to see problems and denied the authority to fix them, three things happen. First, engagement drops, because the gap between capability and permission is demoralising. Second, your best people leave, because they have marketable skills and less patience for structural friction. Third, the organisation starts running two operating models: the formal one, and the informal one that people build around the formal one to get things done. Those informal workarounds are what Stasko was describing. They are not bugs. They are the workforce's rational response to a system that trained them to see better but did not let them do better.

Reciprocity: does the exchange run both ways?

Telstra asked its workforce to transform. To learn new tools. To accept restructuring. To absorb wave after wave of job cuts: from around 32,000 people when T22 began toward 29,000 now, with thousands of roles eliminated in the past two years alone.

In return, the remaining workforce received more capable tools, more training, and less authority. The exchange has a visible imbalance. The organisation keeps asking for adaptation. The structural investment in the people doing the adapting, specifically their ability to exercise judgement and make decisions, has not kept pace.

The A$700 million Accenture joint venture is part of this picture, and it deserves its strongest reading first: specialist capability, speed and shared delivery risk are all legitimate reasons to bring in external help. But the arrangement raises the reciprocity question the public record cannot answer. Telstra spent six years hiring and training a digital workforce to build exactly this capability. What does a A$700 million external partnership to deliver the AI roadmap return to those people: authority, mobility and a credible role in the work, or a message about where the organisation places its confidence? Public evidence cannot establish how Telstra's engineers read that signal. It is precisely the kind of signal an organisation should want to measure before it travels on its own.

Alignment: does what people do match what the organisation says matters?

Telstra says it is becoming an AI-native, digitally transformed technology company. Under pressure, it restructures divisions because of manual workarounds. As recently as February 2026 it could not name a single role that AI had replaced. Its 5G targets vanished between strategy versions without acknowledgement.

The misalignment is not between intention and reality. Telstra's leaders clearly intend to transform. The misalignment is between the pace of narrative change and the pace of structural change. The strategy moves faster than the operating model. Each time the strategy is renamed, the gap grows, because the new story creates new expectations while the old structure remains underneath.

This is visible in what gets dropped under pressure. Not the AI investment. Not the profit targets. What gets dropped is the structural reform, the hard, slow, politically expensive work of changing how decisions get made, how authority flows, and how processes are redesigned from the ground up rather than optimised at the surface.

Signal integrity: does information survive the journey?

The most telling signal in the Telstra story is Brady's February 2026 response when asked about AI replacing roles. She could not point to a specific example. This is not a criticism of Brady personally. It is a measurement of information architecture. If the CEO of a company that has deployed 380 AI use cases and cut thousands of roles cannot connect those two facts with a specific example, something is breaking in the information flow between operations and the executive layer.

The 5G target disappearance tells a similar story. Targets that were central to T25's investor narrative vanished without explanation. Somewhere between operational reality and strategic communication, the signal was lost. Not suppressed, necessarily. More likely filtered by the same instinct that produces clean strategy rebrands: the desire to present a coherent forward narrative, even when the underlying reality is messier than the narrative allows.

The July 2026 outage put a date on the cost. The defect was known. The fix was declined twice. The signal did not convert into action until it became a national incident. Signal integrity failures are abstract right up until the trains stop.

What this means for your transformation

Telstra's pattern is common precisely because it does not look like failure. The profits are up. The AI deployments are real. The strategy decks are polished. If your organisation is mid-transformation, here is how to check whether you have the same gap between narrative and structure.

1. Ask your frontline teams this: "What have you been trained to do that the system does not let you do?" Do not ask leadership. Ask the people who completed the training, received the tools, and then went back to their desks. If they describe workarounds, manual processes, or approval chains that neutralise the capability they were given, you have Telstra's problem. You have invested in empowerment without investing in the authority structures that make empowerment real. The coiled spring is already wound.
2. Count your strategy rebrands, then trace what survived them. How many times has your transformation been renamed in the last five years? Each rename is an opportunity to examine what was quietly dropped from the previous version. Pull the targets, commitments and known risks from your last strategy launch and trace each one into the current version: carried forward, publicly closed, or silently gone? If they vanished without a public accounting, your organisation is using narrative refresh as a substitute for structural accountability. The new name does not fix the old problem. It obscures it.
3. Test whether your efficiency gains are structural or cosmetic. Your CEO should be able to answer this question with a specific example: "Name one process that has been fundamentally redesigned, not just made faster, as a result of this transformation." Not a pilot. Not a use case. A process that used to work one way and now works a structurally different way, with different decision rights, different authority flows, or different handoffs. If the answer is a list of tools deployed rather than a list of processes changed, your transformation is optimising the old system rather than building a new one. That is not transformation. That is renovation.

Where is the gap between capability and permission in your organisation?

One live initiative. Defined scope. Human authority retained.

Patterns on display

Named entries from the Divergent Kind friction register, as they appear in the public record.

• Capability Without Authority. People trained to see problems and denied permission to fix them. The coiled spring.
• Strategy Rebrand Reset. Each new strategy name resets accountability for the last one. Targets vanish without a reckoning.
• Narrative Outpacing Structure. The story changes faster than the operating model. Each rebrand widens the gap it describes.
• Capability Bypass. Paying outsiders to do what you spent six years training insiders to do. The open question is what signal that sends, and how fast it travels.

The register is cross-referenced across this series. Several of these patterns will reappear.

The visible coherence exposure

The visible exposure here is thin, and that is itself the finding. Telstra's costs are structural rather than itemised: a A$500 million savings target quietly revised to A$350 million; a A$700 million external joint venture to deliver capability the company spent six years hiring for; one-off restructuring costs of A$200 to 250 million disclosed at the May 2024 reset, with thousands of roles cut in the two years since; and a nationwide outage in July 2026 whose underlying defect had been logged internally years earlier and left unfixed. The gap shows up as spend that buys optimisation instead of transformation. The July 2026 outage adds a context signal rather than a booked cost so far: an ACMA investigation into Telstra's emergency-call compliance is open, with any penalty not yet determined.

The invisible exposure, capability trained but not exercised across a workforce of 29,000, is larger, and it cannot be measured from outside. Measuring it is what a full Coherence Read does.

Sources and method

This read uses publicly available information. The method names structures, dates, public words, and observable decisions. It does not impute motives the public record cannot establish. Divergent Kind's public Coherence Reads examine public information about non-clients. Client work is confidential and is not converted into public episodes.

Sources

• T22 launch: Telstra Exchange, 20 June 2018
• T22 in retrospect: McKinsey conversation with Andy Penn, September 2022
• T25 announcement: Telstra Investor Day media release, 16 September 2021
• May 2024 job cuts and cost target revision: ABC News, 21 May 2024
• Cost target context: The Conversation, May 2024
• Accenture joint venture: Telstra media release, January 2025
• Connected Future 30: Telstra ASX release, 27 May 2025
• FY26 results and annual report: Telstra, year ended 30 June 2026
• FY25 and prior results: Telstra financial results
• FY26 market coverage: Market Index, 13 August 2026
• H1 FY26 result, role reductions and Brady on AI: Information Age, February 2026
• Telstra International restructure: Information Age, April 2026
• WAVE assessment: The Digital Speaker, June 2026
• 8 July 2026 outage, Senate statement: Vicki Brady, 17 July 2026
• Outage reporting: ABC News, 9 July 2026
• The declined fix: Information Age, July 2026
• Missed update and undocumented design change: The Guardian, 17 July 2026
• Senate hearing coverage: The Sydney Morning Herald, 17 July 2026
• Regulator position: ACMA statement

Richard Lipp is the founder of Divergent Kind and architect of the Qualitative Coherence Indicators framework. His operating career spans Apple, Isobar and Jetstar, NAB Innovation Labs, Qantas, and Virgin Australia. He measures whether the human system is ready for the technology.