Decisions that cost money before they earned trust

Paying the offshore desks for fixes instead of closures

In 2022 Torvelle's service desks in Manila and Pune handled about 2.1 million contacts a year. First-contact resolution was 74 per cent and average handling time under nine minutes. But 18 per cent of resolved tickets were reopened within ten days, user ratings were falling, and in 2021–22 we paid A$7.2 million in service credits across our contracts, A$2.4 million of it to one federal agency with 38,000 users that had begun asking whether offshore support belonged in its next contract.

The cause was in our design, not in our people. Agents were measured on tickets closed and handling time. They were not permitted to reset multi-factor tokens or reinstall standard applications, so many calls ended with advice to restart and a closed ticket, and the user rang back. I proposed removing closure quotas, measuring every team on fixes still holding after ten days, giving first-line agents controlled tools for the commonest repairs and placing second-line engineers in the same pods.

Our CFO, Damien Kirkbride, opposed the timing. Handling time would rise, he argued, and we would need more staff on contracts priced around the old productivity. Vikram Rasquinha, who led our Pune centre, objected to mixed pods because his engineers would lose their specialist career path. Both objections were fair. We went ahead because our credits already exceeded what the extra people would cost.

Handling time rose 21 per cent. We hired 170 people across Manila, Cebu and Pune at A$8.6 million a year, and the change cut about 1.4 points from desk margin in 2022–23. By late 2023 reopened tickets were at 6 per cent and user ratings had risen from 3.1 to 4.2 out of five. Service credits fell to A$1.3 million in 2023–24. Manila attrition fell from 41 to 23 per cent, because agents preferred fixing problems to closing them. The federal agency renewed for five years with offshore support retained for its non-privileged work.

I would do two things differently. I would have told clients that our closure figures were about to fall before they saw them fall, instead of explaining afterwards. And I would have trialled mixed pods in one Pune team first; Vikram was right that specialists needed a visible path, and we built it a year late.

Pricing a state health bid at the roster we could staff, and losing it

In 2023 a state health department tendered a seven-year contract for service desk and end-user computing across 64,000 users, valued in its documents at about A$540 million. It was our largest opportunity. The terms were demanding: every role onshore, many of them cleared, a 90-day transition, service credits of up to 20 per cent of the monthly fee and uncapped liability for data breaches.

Our commercial team's intelligence suggested a winning price of about A$72 million a year. Stuart Coultard, our chief commercial officer, and our chief executive both wanted to bid near that figure and find savings during transition. I built the roster instead: a 24-hour onshore desk of 310 full-time equivalents, a field team for 140 sites, and the clearances the contract demanded. At A$72 million those people would have needed to work at 91 per cent utilisation, with no allowance for training, leave or the first year's turnover. A sustainable price was A$83 million.

I recommended bidding at A$83 million, proposing a liability cap of twice annual fees and a 180-day transition, and explaining why. We expected to lose. Damien Kirkbride, our CFO, supported me after testing the roster himself. The board's bid committee approved our price after a long discussion in which one director argued that losing the account would set our government business back two years.

We lost. We had spent A$3.1 million on the bid and planned 210 roles we did not fill. Government revenue grew 4 per cent in 2023–24 against a plan of 9 per cent, and two bid managers left. I do not know whether the winning price was sustainable, and it is not my place to guess.

What changed was our process. Every bid above A$50 million now reaches the board with its staffing plan, clearance timetable and liability position before its price. Our decline rate on qualified opportunities rose from 12 to 31 per cent, our win rate on submitted bids from 29 to 41 per cent, and margin on government contracts held at 12.6 per cent.

What would I change? I would have joined the department's market-sounding sessions a year earlier and argued about liability before the tender was written, when a supplier's view can still shape the terms instead of merely declining them.

Taking standing administrator rights away from offshore engineers

By early 2024, 340 engineers in Manila and Pune held standing administrator rights to client environments. For government contracts that was already restricted, but for enterprise clients it was normal, including three water and energy businesses with obligations under the Security of Critical Infrastructure Act. Australia's cyber security strategy, published in late 2023, and the questions arriving from those clients' risk teams made it clear that they would soon ask who could reach their systems, from where and with what approval.

I proposed that all privileged changes for government and critical-infrastructure clients be made only by cleared engineers in Brisbane and Adelaide, through just-in-time access from managed workstations, with every session recorded. Offshore engineers would continue to diagnose faults, prepare changes and test them outside production. The cost was A$13.5 million a year, mostly 96 new onshore roles.

Vikram Rasquinha, then leading Pune, disagreed strongly. His engineers had met every control we had asked of them, and he read the change as a judgement on their integrity. Damien Kirkbride, our CFO, pointed out that two of the three utilities would not pay more for it. I respected both arguments and kept the decision, because a supplier's access model should not wait for a breach or a client's audit to change it.

We redesigned 70 Pune roles around testing and application maintenance. Eleven engineers resigned within six months, several of them among our best. Gross margin on the affected contracts fell by 1.8 points.

In February 2025 a phishing campaign captured the credentials of three Torvelle engineers. Because none held standing privileges, none of the accounts could reach a client system. We told the affected clients within hours and commissioned an external review, which found the access model had worked as designed. Later that year Kelvanne Water Corporation awarded us a six-year, A$96 million managed-infrastructure contract, and its chief information officer told us our access model was a deciding factor.

I would do the human part differently. I announced the change on a video call and travelled to Pune a week later. I should have gone first, explained the client risk in person and shown Vikram's team the new career path before anyone heard the word restriction.

Lowering the field engineers' utilisation rate after a crash

In March 2025 one of our field engineers, finishing a thirteen-hour day that included two regional callouts and a late change window, rolled his vehicle on the highway east of Roma while driving home. He was seriously injured and has since returned to work. Our investigation found that he had broken no procedure. He had done what our dispatch system and our utilisation rate asked of him.

Field engineers were planned at 88 per cent billable utilisation. In regional areas, where travel and waiting for parts consume much of the day, the only way to reach that figure was to stack callouts and work late. Overtime averaged 31 hours per engineer a month, median annual-leave balances had reached seven weeks, and annual attrition was 27 per cent.

I proposed lowering the rate to 78 per cent, hiring 60 engineers, capping combined working and driving time at twelve hours and repricing regional callouts with clients. Craig Tolhurst, who runs field services, supported the safety changes but argued the utilisation cut was too deep and would make his business unprofitable on three contracts. Damien Kirkbride calculated that each point of utilisation was worth about A$1.4 million of field-services EBITDA. Both were right about the arithmetic. I asked the People, Safety and Remuneration Committee to approve the change and to remove utilisation from my own incentive measures, which it did.

Field-services EBITDA margin fell from 13.4 per cent in 2024–25 to 10.9 per cent in the half to December 2025. We repriced regional callouts on two large contracts and lost one smaller client worth A$6 million a year that would not accept the new rates. Margin recovered to 12.2 per cent in the half to June 2026. Attrition fell to 14 per cent and overtime to 12 hours a month. The lost-time injury frequency rate fell from 6.1 to 2.4, and regional priority-two callouts met their service level 92 per cent of the time, up from 81.

The board now sees field hours, consecutive on-call nights and leave balances beside utilisation every quarter.

I would have made this decision a year earlier. The overtime and leave figures were in our reports throughout 2024, and I read them as a productivity question when they were already a safety one.