Companies have myriad software development strategies to choose from, but it wasn’t the completion of a recent AI-driven project in half the expected time that convinced the head of Woodside Energy’s data strategy that they had found the right approach.

Rather, explained chief data officer Shelley Kalms, that validation came when another project was not delivering the value that had been expected – and the project was stopped in its tracks.

“We put the initiative up and it got approval and funding, and the squad stood up,” Kalms told the recent IBM Think conference, “and they started working towards a hypothesis. But what they were finding was that it wasn’t delivering the value they had forecasted – and so they stopped.

“You could say that it is failing, but it’s also learning in that place – and the team learned that ‘we’ll go fast, we’ll determine are we delivering value?’ and if we’re not, then we stop.”

The ‘fail-fast’ mantra is nothing new for startups that often live or die by it, but within larger organisations it can still be a surprise to see it actually happen – hence Kalm’s delight at seeing the philosophy embraced so effectively.

Stopping an underperforming project, after all, allowed Woodside to redirect its valuable data analysts, developers, AI specialists and other experts to other, more beneficial projects – such as a new, AI-based platform that has proven particularly effective at predicting when the resources giant’s machinery needs to be serviced.

Development of that system was originally kicked off with the expectation of a 12-week implementation to develop a minimum viable product (MVP) – but “they actually did it in 6 weeks,” Kalms said, “and they started to realise value that went to the P&L [profit and loss statement].”

“That to me was real,” she added. “We’re working at the speed of a startup but we’re scaling for an enterprise.”

Learning to fail

Woodside’s commitment to digital innovation stretched back to at least 2014, and the company is often cited amongst Australia’s most progressive due to its full embrace of data-driven transformation and operation.

The company’s partnership with IBM, in particular, has fuelled a range of projects including an AI-powered, risk-based maintenance system and an HR onboarding tool that was run as a test of the evolving IBM Garage collaborative development process.

Helping enterprises function like a startup – and yes, that does include failing fast – is a core mission of IBM Garage, a self-professed “transformation accelerator” that leans on virtual collaboration to bring together company developers and IBM specialists in AI, blockchain, data science, and other critical areas.

Although it speaks the language of startups, IBM’s collaborative approach targets a chronic problem: big companies still aren’t very good at managing data.

A recent MIT Technology Review Insights survey showed just how badly data is being managed in most companies, with just 13 per cent of companies realising their data-based goals and many others floundering.

Successful data projects require more than tools, with successful corporate data users reiterating the importance of a data-focused culture as well as the tools and the data itself.

Garage is as focused on building that culture as it is on improving access to tools.

That has well suited Woodside’s forward-leaning agenda, with collaborative teams brought together and working together regularly to remove internal friction – ensuring that the project has access to the right skills, when they’re needed.

“Culture is absolutely key to transformation,” Kalms explained, noting that the Garage model “takes that human-centric approach from all aspects, when we’re looking at the end-to-end.”

“It takes the research and takes it in the context of where it sits. The data then drives the solution to the opportunity or the problem that you’re trying to solve.”

There’s always a catch

It’s not always a home run: at many companies, Gartner recently observed, greater reliance on agile development and remote collaboration have been a mixed bag.

While 92 per cent of agile teams boosted their code output by 10 per cent during the pandemic, Gartner noted, two-thirds of software teams are releasing code less frequently – and it’s 64 per cent larger, on average.

“This is not the ideal situation for agile, newly remote teams,” the analyst firm notes.

Part of the challenge of an effort like IBM Garage, therefore, is not only facilitating collaboration but keeping it from taking on a life of its own.

For Woodside, Kalms said, widespread buy-in has been key to keeping its projects from bloating out of control.

“People come to this on day one,” she said, “and they are looking at the concept, the ideation, the implication and run, and they’re seeing it all the way through”, even if that failure means ending a project prematurely.

“What’s different is that it’s not being done to us,” Kalms said. “It’s our people that are developing a solution that is extremely engaging and enabling, and that takes it to a level I haven’t seen before.”