 in January, it didn’t make as many headlines as it might have, being overshadowed by DeepSeek’s LLM launch and the early signs of what Trump’s comeback means for the world. But beneath the radar, it was a significant step forward for consumer AI Agents, an area where many startups and investors are betting big.

It seems clear that AI will fundamentally reshape the way we all live our lives, but ‘AI’ is such a broad and generic term that it’s almost meaningless when it comes to deeply thinking through what those changes might look like and what opportunities it might create for startups. The problem, though, is that the more specific you dare to be, the more likely you are to be completely off the mark. As the saying goes: 

“It is better to be vaguely right than exactly wrong.”

We disagree. When things are changing so quickly and we’re continuously developing our thesis on how things might play out, we think it’s better for our work-in-progress opinions to be “exactly wrong rather than vaguely right.” Teasing out the nuance helps us ask better questions of ourselves and founders without getting deafened by generic narrative and buzzwords. We believe coupling a comfort with being “exactly wrong” and a willingness to evolve our views based on strong counterarguments helps us iterate more quickly towards a better perspective.

With that said, we wanted to lay out some initial thinking on AI Agents in consumer journeys to prompt discussion.

Our oversimplified (work-in-progress) categorisation is this:

Already, that’s clearly flawed. The reality is that there is a spectrum of Assistant to Agent within different products and consumer journeys.

On the Assistant end of the spectrum, we think about something like ChatGPT. To be overly reductionist, AI Assistants are really about better Search. This is already a massive deal: evolving from “a list of potentially useful links for people like you” to “the most likely best answer for you” is a stepchange - 

 in just over two years from launch isn’t bad! But it remains a fundamentally user-directed journey.

At the opposite end of the spectrum sits some of the ‘futurecasting’ speculation you hear from the most excited advocates, which broadly boils down to: “AI will know everything about you and solve everything for you without lifting a finger.” Do your groceries. Book your holidays. Fill your wardrobe with things that make you look more like Chris Hemsworth. And so on to utopia.

In the messy middle sits Operator’s current apparent level of sophistication. It can take action on your behalf (e.g. pre-filling your basket on Instacart, or finding the highest-rated tour in Rome) but it doesn’t (yet) complete the end-to-end transaction. It’s an Assistant with some agency, so definitely 

agentic than ChatGPT, but not an Agent in its strictest sense. At least for now.

A key question we’re debating is: when will consumers be more likely to trust an AI Agent to take action on their behalf, rather than just assist? As exciting as the prospect of outsourcing large portions of our lives to AI Agents might be, there are clear reasons why adoption may be slower than some expect. Here are some key factors that could hold them back:

Not all decisions can be boiled down to clear, structured instructions. For visual categories like furniture, apparel and art, the experience of looking through the options isn’t just about finding a match within a well-defined criteria, it’s about 

the criteria to begin with. If we don’t know what we want until we see it, can an AI Agent truly deliver the best outcome?

Some types of shopping are part function, part enjoyment. Browsing for travel destinations, discovering new fashion brands, or exploring niche wines isn’t just about finding the “best” option - it’s about the process of exploration and discovery. AI Assistants can help narrow down choices, but will users want to miss out on big parts of the fun?

Too much curation can backfire. Consumers want guidance, but not at the cost of feeling like they’re missing out on options they would have explored themselves, particularly within categories they know well or have strong opinions on. If I trust my own research and instincts, will I really delegate an important choice to an AI Agent?

4. If There’s a High FOFU (Fear of Frustrated User)

The more complex the request, the more opportunities for an AI Agent to get it wrong. 

Much harder. Over time, AI Agents will improve at understanding user preferences, but will consumers tolerate the inevitable mistakes along the way or just opt to do it themselves?

Decision-making isn’t always consistent. A business traveler might prioritise comfort and convenience over cost, while the same person might be price-sensitive when booking a family holiday. Will AI Agents have enough context to accurately adjust for these shifting priorities without needing constant micromanagement?

Some choices rely on emotional or subconscious factors - things that are hard for AI to quantify. The difference between two restaurants with similar ratings might come down to ambiance, personal memories, or a brand’s identity. Can AI really make 

 judgment calls that feel right as well as being the logically ‘optimal’ choice?

Trust is a major hurdle. Today, we know search engines are ad-driven, but we also know we can scroll past the paid results. AI Agents shift the paradigm from 

 Without transparency on who is paying and how, will users believe the AI Agent is truly working in their best interest rather than selling them to the highest bidder?

As in any context, there’s a risk of over-delegation for stuff that really matters. Taking the plunge on an AI Agent completing tasks where the monetary, time and/or emotional stakes are low makes sense. But when it comes to more important things like legal matters or personal finances, having the human in the loop as the ultimate decision-maker is a feature, not a bug.

Many of the above factors boil down to a simple equation inspired by 

Scott Galloway’s “algebra of disincentives”

. In short, true agency is more likely if:

Probability of Bad Outcome x Severity of Bad Outcome < Hassle Factor of DIY 

In other words, the adoption of Agents will be limited to contexts where the downside risk is less than the hassle of a user doing it themselves. The ‘Probability of Bad Outcome’ will drop over time as technology and context improves, but we’re skeptical it will reach zero across the board.

To illustrate, here’s a rough framework we’ve been playing with for where AI is more (or less) likely to be truly agentic for certain tasks:

It’s not perfect, by any stretch: there is likely plenty of ‘exactly wrong’ in here. But as a general framework to kickstart our thinking, we’re finding it pretty useful.

AI Agents are coming, but their success will depend on more than just capability. Consumer behavior is driven by psychology, trust, and habit. The best early use cases will be simple, transactional, and low-risk. Over time, as trust builds and models improve, we may see more consumers hand off complex, higher-stakes decisions. But we’re not there yet.

Where do you think AI Agents will break through first? Tell us where we’re exactly wrong.

And if you’re a founder tackling these problems, we’d love to hear from you!

AI Agents show promise but face adoption challenges, with success likely first in simple, routine tasks where downsides are minimal compared to the hassle of doing it yourself. Joe's blog explores the spectrum from AI Assistants to true Agents, examining when consumers will trust AI to act independently rather than just assist.

Consumer AI Agents: Ready for take off?

The womenswear launch comes at a strategic time, with 20% of its current customer base being women and many of its partners requesting a womenswear collection. To design and develop the collection, Represent brought in former Topshop head of design Toni Purdie. Purdie leads an all-female design team and brings experience from Stella McCartney, Acne Studios and Burberry.  

The collection is centred around “everyday Icons” - versatile pieces that balance masculine and feminine elements while maintaining the brand’s commitment quality and sustainability.  

 already being stocked in over 150 retailers all over the world, including big name department stores such as Selfridges, Harvey Nichols and Saks Fifth Avenue, the brand recently opened its flagship store in its hometown of Manchester.  

Represent’s womenswear launch debuted February the 10

 with a Selfridges pop-up, with the collection launching online soon. 

As hinted at the end of last year when we announced our strategic partnership with Represent, the brand has now expanded into womenswear. Founded by brothers George and Michael Heaton, Represent has a loyal customer base and has seen considerable growth.

Investment, reimagined

Private Equity Investments up to £100m

Venture Capital Initial Investments up to £2m

Apply here

Office Hours

True Investment stories

Insight

Consumer AI Agents: Ready for take off?

News

Represent launches womenswear