Published on the 14/09/2023 | Written by Heather Wright

The AI pillars Gartner says you need to nail now…

AI for productivity is merely tablestakes. The true value of the technology will lie in ‘game-changing AI’ – but that AI comes with some big caveats, CIOs at Gartner’s IT Symposium in Australia were warned this week.

Unsurprisingly, AI was the topic of this year’s Symposium, with hardly a session passing without mention of the topic. The entire keynote of the conference was devoted to the topic – the first time the opening has been focused on a single topic.

“Your job is to be the executive AI guide, guiding the executive team to explore your AI opportunities and risks.”

With AI at the peak of its hype cycle, Gartner distinguished VP analyst Don Scheibenreif and Neha Kumar, senior director of research and advisory, urged attendees to move quickly to capitalise on the opportunities, while not falling for the hype.

The two stressed that Generative AI is part of a larger shift in how machines and humans interact, with machines moving from being our tools, to being our teammates.

“This is a story still being written and you, the CIO, have a big role in how things play out,” Kumar told attendees.

Survey results show Australian and Kiwi tech leaders are both excited and cautious, about AI: 61 percent say AI is the top technology for innovation, with 51 percent flagging generative AI as tops. But just 33 percent believe their organisation can mitigate the risks – well below the global average of 47 percent.

Kumar and Scheibenreif were clear that it’s in the hands of CIOs to set the path through being intentional about what they want from AI. 

The two analysts outlined an ‘AI Opportunity Radar’ which they encouraged all companies to work through, deciding with the wider enterprise what AI ambition an organisation has, which zones they will play in and which they will avoid.

The AI Opportunity Radar splits AI into four ‘opportunity zones’, with internal and external facing offerings for both the everyday AI half – focused on productivity – and the ‘game-changing AI’ half, which Scheibenreif dubbed a ‘reinvention play primarily about creativity, not productivity’.

“Public or private sector, now is the time to put this AI opportunity radar in front of your executive team and decide where you will and will not play,” Kumar says. 

Scheibenreif says everyday AI is where 80 percent of A/NZ organisations are focused right now, in line with the global average. On the internal side there’s the use of GenAI to write code better, or to augment your IT Helpdesk or have strategy teams use AI to put together SWOT analyses so they can spend more time analysing data than collecting it. 

On the external side, there’s front office options, such as generating content for external communications.

“Choose your AI ambition and fill up the left-hand side of the radar with your own everyday AI opportunities and identify everyday AI opportunities,” Kumar urges.

But Scheibenreif says everyday AI is the new tablestakes. 

“Everyday AI will go from dazzling to ordinary with outrageous speed,” he says.

While it might feel dazzling for every department – and will bring ‘remarkable’ results, it won’t give an organisation a sustainable competitive advantage because everyone has access to the tools. 

Game-changing AI will reshape create and even destroy industries, he says.

The pair outlined several examples where AI is already bringing game-changing results, including in drug discovery, where Hong Kong biotech Insilico’s Pharma.ai were able to nominate nine drugs last year, several of which made it to phase one clinical trials – well ahead of the usual four to five nominated a year by big Pharma.

Gartner has forecast that by 2025 more than 30 percent of new drugs and materials will be discovered using GenAI, with AI enabling drug discovery to move beyond big Pharma to smaller companies. 

Khan Academy, meanwhile has debuted the Khanmigo AI powered teaching guide, providing an interactive virtual tutor for students and enabling them to ‘chat’ with, for example Marie Curie about radioactivity, or Gandalf about Lord of the Rings.

But Kumar warned that game-changing AI comes with a health warning: A high price of entry and high risk of things going wrong.

“To do game changing AI three tough conditions have to be met: You will need lots of risk tolerance, a lot of executive patience and boat loads of money,” she says, noting those are the exact same conditions driving digital business transformation.

With CFOs already unhappy about current digital investments – 67 percent of CFOs say digital investments are underperforming – you’d expect that wrangling that extra cash could prove challenging. But 66 percent of A/NZ organisations (and 73 percent globally” say they plan to increase spend on AI.

The pair outlined three investment scenarios for local companies: Defend, extend or upend.

Defending is back to tablestakes – investing in quick wins that improve specific tasks, such as using Microsoft Copilot.

Extend gets more expensive, but also more valuable, investing in custom application using public data and your own data.

Upending your organisation and disrupting your industry to pursue new products and business models powered by AI is where things get ‘really expensive, really fast’, Kumar notes. 

“But it also comes with a much higher reward. It could be amazing.”

Nonetheless, it’s an area Gartner isn’t expecting many A/NZ organisations to play in because of the risk and time investment required. In fact, the analyst firm has forecast that by 2028 more than 50 percent of enterprises that have built their own large models from scratch will abandon their efforts due to costs, complexity and technical debt.

The four disruptions

Scheibenreif and Kumar also urged companies to consider four disruptions anticipated:

• AI will make something free in your industry
• AI will make something obsolete as in education when students can chose AI tutors over human teachers
• AI will create a net new need that wasn’t there before, just as step counting and sleep monitoring proliferated in recent years
• And, for the private sector, AI will cause competitive barriers to crumble.

“Discussing these disruptions is your and your executive team’s responsibility,” Kumar says. “It will help you prepare for the coming shockwaves and inform where you place your bets.”

“Game-changing AI means big disruption,” Scheibenreif says. 

“Your job is to be the executive AI guide, guiding the executive team to explore your AI opportunities and risks. 

“You need to decide on your optimal investment porfolio – are you going to defend, extend or upend your industry?

“And brace for shockwaves. The way to do this is to challenge your long-held beliefs about what humans should do and what machines could do.”

Three pillars for CIOs to nail

So what’s a CIO to do next in order to be AI ready?

“There are three pillars the CIO has to nail: AI-ready principles, AI-ready data, and AI-ready security,” Kumar says.

Even staying firmly in the everyday AI realm, doesn’t mean everyday risks, with Kumar warning that everyday AI is where people will run into human to machine challenges first. 

She recounted how her son will smile and ignore her when she tells him it’s time for bed, but will go to bed when Google says it’s time. It’s a situation Google’s developers likely didn’t expect when creating the technology.

“In this new realm of human to machine interaction where we talk to machines and they talk to us and we listen, there will be all sorts of unforeseen consequences. 

“What this means is you need to think ahead of time about what lines you want to cross. 

While regulators around the world are working to create some of those lines, the lack of regulation is causing hesitation in AI use for 37 percent of A/NZ organisations, and 47 percent globally, according to Gartner Research.

“To move forward you need lighthouse principles that light the way even when everything is new, murky or unclear,” Scheibenreif says.

Those principles are driven by your organisation’s values – but just five percent of A/NZ organisations have an AI vision statement in place. 

“Lighthouse principles are not generic, they are not platitudes and they are never ambiguous,” he says. 

And they are critical, he adds.

“Take vendor selection. When you are buying AI software you are not just buying technology. It’s like you are hiring a teammate. Is that teammate going to steal your enterprise data and throw it on the internet, or is it going to have your back?

The principle here would be to involve HR every time you acquire user-facing software.”

There was some good news on the data front.

While chances are your data isn’t AI ready – 96 percent say theirs isn’t – not all data needs to be AI ready.

“We’ve been taught to think there is mountains of enterprise data, mountains of gold but lot of that data is fools gold, not that useful,” Kumar says.

“It’s your proprietary data – your algorithms, formulas, blueprints schematics – that’s the real gold.

So you don’t have to make all your data AI ready. It’s just the stuff that serves your AI ambition.”

Getting it ready means data is secure, enriched, free from bias, accurate and governed by your lighthouse principles. 

And enriched data means data plus rules plus tags, making that data ready for large language model consumption.

Providing rules means for example, robots in warehouses don’t just need data, they need to be taught the laws of physics so they can move around safely, while for AI to help lawyers the machine needs to be taught the rules of law. 

“This means you won’t necessarily need massive data sets. A small amount of data accompanied by the attendant rules might be enough,” Scheibenreif says. 

On the tagging front, he says meta data is almost as important as the data itself.

Meanwhile, AI opens up many new attack vectors, both direct and indirect.

Scheibenreif says LLM grounding, which reduces the likelihood of generating answers that drift from being accurate to inappropriate can help with direct user threats, while for indirect prompt injection existing security tools can help. 

“Hijacking inputs isn’t a new thing, it’s just accelerated with AI,” he notes.

“As AI security risks evolve and emerge you will have many tools at your disposal and you’ll just do what you’ve always done: Get to know the new attack vectors and then prioritise investment to address them.”