November 20, 2026 · 5 mins read
"Using AI didn't make it cheaper. It made it braver." What we learned at Interact 2026
Most teams now use AI every day. Fewer can say what it should do, and what it shouldn’t.
That gap is why we ran Interact 2026. GAIN started Interact in 2017 as a UX research event, and more than 5,000 people have joined since. AI has broken down the walls between design, data, experimentation and marketing, so this year we widened the conversation. Under the theme “Designing tomorrow”, we looked at what AI makes possible, where it goes wrong, and where people still make the difference.
We brought five sessions together at Fora, South Bank, with our partners Acquia and Wingify:
- The last 10%: Chris Richards, Creative director, GAIN
- Drop the “ING”: Jennifer Griffin Smith, Chief market officer, Acquia
- From consideration to conversion: John Hughes, VP of marketing, Wingify
- Panel, From theory to practice: Mark Williams (David Lloyd Clubs), Steph Wright (Our AI Collective) and Alastair Campbell (GAIN), chaired by Peter Kay (GAIN)
- Everyone is a manager now: Didem Ün Ateş, Founder and CEO, LotusAI
Below is what each speaker said, a preview of the slides and one thing you can try with your team this week.
1. Spend what AI saves on the last 10%
Chris Richards, Creative director at GAIN, opened with two chairs. A Victorian desk chair was made to order and built to last. The bentwood chair that followed was quick to make and became the template for most chairs since. AI gives us the same trade-off. It can do 90% of the work, but people connect with the last 10%: the craft and emotion.
Coca-Cola's AI Christmas ad tested well, then lost goodwill once viewers knew how it was made. Prada went the other way and hired animators to hand-draw its campaign. In GAIN's own work, AI prototyping cut a two-to-three-week process to about three days. The budget stayed the same, so the team could test bolder ideas. As Chris put it: "Using AI didn't make it cheaper. It made it braver."
Try this: put your brand, services and past work in one place before you prompt, so AI starts from your material. And remember Chris's last line: to be good at AI, you need to be just as capable without it.
2. Ask AI about your brand before your buyers do
Jennifer Griffin Smith is Acquia's chief market officer, not chief marketing officer. She dropped the "ING" on purpose. Marketing has become the team for producing and publishing, just as CMOs are asked to drive growth. AI can take on that work, from translation to resizing. But it also amplifies bad content. Acquia's research found 61% of CMOs have no single source of truth for their content.
She left us with five things to do, none of which need budget sign-off:
- Ask AI about yourself. Run ten real buyer prompts. See what comes back and who gets cited.
- Fix before you scale. Audit your top 20 pages for structure, accuracy and proof.
- Book your CIO. Put a standing joint roadmap review in the diary.
- Pick one shared number. Add one customer outcome to both scorecards.
- Measure answers, not traffic. Track AI visibility and citations next to rankings.
3. Most ideas lose, so test before you trust them
John Hughes, VP of marketing at Wingify, started with a game: colour logos or greyscale logos under a demo request form? Greyscale won, with 33.93% more demo requests, because the colour logos pulled attention away from the form. Most ideas don't win. Even at Google and Bing, only 10 to 20% of experiments produce a positive result.
Buyers now research with AI before they reach your site, so pages need to work for machines and people. On personalisation, keep a holdout: a small group who still see the original page. Without one, you can't tell if it's working. In John's words, "personalisation without a holdout is a guess with a bigger budget".
Try this: next time you personalise a page, keep around 10% of visitors on the original and compare the results.
4. Just because you could, should you?
Our panel, chaired by Peter Kay (GAIN), started from that line from Jurassic Park. Mark Williams showed what doing it properly looks like. David Lloyd Clubs spent 18 months building an AI recommendation engine for members. Sports scientists and doctors set the rules. Steph Wright of Our AI Collective reminded us that "FOMO is not a strategy". Trust has to be earned. Alastair Campbell (GAIN) explained that AI learns from an inaccessible web, so its output is often inaccessible too. Build accessibility in from the start.
Try this: before your next AI project, ask who it could leave out, and how you'll earn your customers' trust.
5. Everyone is a manager now
Our keynote speaker, Didem Ün Ateş, founder and CEO of LotusAI, helped set up Microsoft's Office of Responsible AI. Her view: AI's risks can be contained if it's done properly. As AI agents take on more work, everyone becomes a manager. In a 2026 Wharton survey, 72% of managers said their focus has moved from creating work to managing AI's output. That takes four skills: judgement, delegation, quality control and accountability.
Try this: set three to five success metrics before you start, and measure them. As Didem put it: "Who goes on a journey so expensive and frustrating without success metrics?"
The thread through every session
One idea ran through every session. AI can take on more of the doing, but people still need to judge, check and decide. Thank you to our speakers, our partners Acquia and Wingify, and everyone who came along.
For more from the day, keep an eye on thisisgain.com and follow GAIN on LinkedIn.
Recent Posts
See All-
The video that cost WeBoost 10% of its sales (and led to a 46% win)
The video that cost WeBoost 10% of its sales (and led to a 46% win)
-
When Good Ads Go Bad: The Hidden Cost of Creative Fatigue
When Good Ads Go Bad: The Hidden Cost of Creative Fatigue
-
Left it until November? Your ads can still win you Black Friday.
Left it until November? Your ads can still win you Black Friday.