The GamesIndustry.biz HR Summit took place on October 1, and one topic that came up again and again at sessions throughout the day was generative AI and the impact it’s having on the workplace.
Attitudes towards the technology were mixed. Some pointed to how AI has been able to speed up some aspects of the game creation process, such as coding, or automate drudge work. But the general view was that generative AI falls far short of the miracle cure that some AI enthusiasts pitch it as, and that in addition it has been causing all sorts of workplace problems as firms figure out how best to use and account for it – including worries about the privacy of sensitive company data, an increase in employee anxiety, and a higher workload for HR professionals as a result of lengthy AI-written grievance letters.
One senior leader emphasised that their company was meticulous about not letting any AI-generated content end up in a finished game (the HR Summit ran under the Chatham House Rule). But at the same time, their company recognises AI might have uses elsewhere. “So it’s a fine balance,” they said, acknowledging that “there could be good opportunities for efficiencies and doing some of the more routine jobs or mundane jobs that we never want to do in game development.”
Yet there remains huge anxiety among employees about the use of AI, they warned: “I’ve had emails from people at nine o’clock at night going, ‘Are we using AI? Oh my God, does that mean my job’s going to go?'” That anxiety extends to management teams, which worry about where the tech is going and how to plan for it. “The reality is that we don’t know what the future is going to look like,” said another senior leader.
At last year’s HR Summit, a show of hands revealed that only a minority of attendees’ companies had an AI policy in place. This year, a similar straw poll revealed that most now have an AI policy. But the sheer speed of change is causing headaches. “I wrote a policy two weeks ago, and then I found out about some prompt injection yesterday that means I’ve got to go back and rewrite some of the guidelines,” said one company leader. “It’s changing that quick.”
One of the biggest worries is sensitive data leaking outside the company. Even if firms pay for an enterprise version of an AI model that ensures none of the company’s data is used for training by the AI firm, they have little control over employees’ individual use of AI models. “They could be going off to ChatGPT on their personal account and putting all sorts of NDA things in there. And I don’t see how you can physically stop people doing that,” said one leader. “So being a bit clearer about what we think is acceptable and unacceptable, that’s the best way forward.”
“I haven’t seen AI create anything fun”
Yet there’s still confusion about exactly how AI should be used. “The biggest issue at the minute is that AI is the answer, but people don’t quite know what the question is,” said one leader, who described multiple studios seeking to “use more AI” without a clear idea of what they want it to do.
“We’re in this kind of over-hyped world. We’re all hoping that this great panacea is going to open up, it’s going to be amazing. And like every other advance, it’s not.”
One senior-level attendee summed it up bluntly: “I haven’t seen AI create anything fun. It can replicate, for sure, but I haven’t seen anything fun come from AI yet.”
“You can make crap quicker,” quipped another.
AI workflows
Whatever your attitude towards AI, the message at the HR Summit was clear: you can’t ignore it. Even if your company doesn’t want to use generative AI in any form, that needs to be enshrined in employee guidelines and enforced, and everyone needs to keep up with what’s going on.
“It’s really about understanding what the current state of AI technology is, understanding how that aligns with the values of your organisation, how it aligns with the people of your organisation, and the risks and opportunities that then creates,” said one speaker. “AI technologies are now encroaching on a lot of creative disciplines that five years ago we did not even consider would be an issue.”
Yet we’re only really at the start of the generative AI journey. One session outlined the four stages of AI adoption, from dabbling with AI to becoming an AI-first company. The vast majority of companies are still at the first stage, perhaps buying an enterprise license for Claude to let programmers play around with generating code. But a lot of larger studios are in stage two, where they are trialling generative AI in certain workflows, evaluating the time savings against the potential cost and risk.
“There were more fingers than most humans have on this particular piece of art”
Without rigid enforcement, however, chaos can ensue. One attendee provided an anecdote about an artist using Gemini to generate some shader code while a programmer was out of the office. The inefficient code, which didn’t conform to the company’s coding standard and went against the studio’s terms and conditions on generative AI, was only spotted weeks later, and resulted in the build being reverted.
Another anecdote involved outsourcing a piece of art to a freelancer, who used AI to create it. “There were more fingers than most humans have,” said one speaker. “But another person outside of that department went, ‘That looks great, ship it’, and the community went absolutely wild. So that time and cost saving they thought they got wasn’t there.”
AI workflow pilots can also reveal fatal flaws in studio processes. “Most of them collapse, because [workflows are] driven by the tacit knowledge of individuals rather than formally documented operating procedures,” said a speaker.
There’s also the human side to consider. Even if an AI-driven workflow allows programmers to submit more code more efficiently, they said, “we also have to look at it from the qualitative side. Yes, you’re able to move faster, but did this improve the quality of the process for the people involved? Because a lot of the times we move faster, but it was a miserable experience: ‘I spent most of it just shouting at an AI prompt thing that I couldn’t quite get to do what I wanted it to do’.”
HR headache
From an HR perspective, a few of the summit’s attendees pointed to how generative AI can be beneficial. In terms of writing emails, “sometimes AI is really great for helping people frame up what they really want to say,” said one attendee. “Sometimes you just need a bit of a prompt or a kick just to get your thoughts in order.” AI-based HR systems can also provide automatic answers to commonly asked questions about things like maternity or sickness leave.
But AI is also generating more work. Several people pointed to an increase in employees using generative AI to write grievance letters, which can be many, many pages long. One legal representative said they receive around ten such letters from clients every week. “They generally go long, with quite a lot of factual detail that isn’t quite correct, and often the legal references are quite odd.”
“Eight pages of it contained things that we couldn’t even make sense of”
The problem is that HR representatives are dutybound to faithfully read and respond to these lengthy AI-written grievances, even if they cite incorrect information or make irrelevant claims. “It’s not doing any favours for the claimant either,” said one attendee. It was noted that if the employee can’t show that they understand the details of the AI-generated report, the case could be thrown out by a tribunal.
“We had one not long ago, and it was a grievance an employee submitted, a formal one, about nine pages long – and probably eight pages of it contained things that we couldn’t even make sense of,” said one attendee. “Bear in mind, it’s just the first stage of the formal grievance on something that was actually relatively minor that could have been sorted out between myself and the line manager and the employee.”
Yet the solution isn’t as simple as simply banning AI-authored letters of complaint – as one attendee pointed out, some employees might need support with writing, and banning AI use could be seen as discriminatory. But another attendee suggested a potential solution of sorts by adjusting the company’s policy to stipulate a maximum word count for grievance letters, or responding with a request for specific timelines, dates, and details that an AI wouldn’t know, rather than spending hours with the legal team trying to decipher the complaint.
On balance, it seems like generative AI could actually be generating more work for HR teams. “The reality of it is that we’ve had HR teams that have shrunk, but the amount of work that you have has stayed the same or gone up,” said one attendee. “And just because people are using AI or using LLMs to try and help these things, actually quite often it’s given me more to do.”
“It doesn’t have heart. It can’t read the person”
They pointed out that providing AI-generated replies to common HR questions might mean HR representatives are missing the essential human connection with the employees they’re charged with helping. “Because the person who asked about the maternity policy might be having to go through IVF and they need some additional support, but they just found the answers out without realising there are a million other things that we can do for them as well. They’re asking about the sickness policy because somebody needed to go and have an operation in a hospital, and they were worried that they weren’t going to get paid for the time that they needed for the recovery.”
The AI can answer the question, but it can’t provide empathy or pick up on emotional cues, they concluded. “It doesn’t have heart. It can’t read the person. It can’t tell what else might be going on. It can’t feel the vibe in the room when you walk in, or see someone’s body language.”
“That connection with people is the glue that brings it all together.”