Are you sure you want AI matchmaking in your event app?

AI matchmaking seems to be having a moment.

You know the kind of thing I’m talking about: a feature within an app that proports to use machine learning to suggest the most valuable people for you to meet at the event.

It seems like barely a week goes by without someone – usually a customer (existing or prospective) asking me if AttendZen offers this as part of our attendee app.

My response is to smile benignly and say we don’t.

Usually, whenever customers express an interest in, or enthusiasm for, a feature we don’t have, my natural instinct is to tell them about products we integrate with, or our work-around, or where the missing function sits on our roadmap.

But not in this case.

No, the reason we don’t do AI matchmaking isn’t because it’s difficult or expensive to engineer (it’s not).

We don’t offer it is because we take the view that it doesn’t work and its main contribution to most business events is simply to annoy people.

If you’ll indulge me, here are my problems with this kind of ‘forced networking’, and some thoughts on better approaches you can take instead to promote meaningful human connections at your next B2B event.

AI matchmaking over-promises

Let’s start with the hype.

AI networking tools are sold to us as being ‘intelligent’. The pitch is that they somehow understand human professional needs better than humans themselves. Clever algorithms mine your participant data for buried insights, like the notion that Gerry in Sales would probably like to connect with Marco in Purchasing. Ya think?!

Look under the hood though and these ‘intelligent’ apps actually function more like basic filters. And really that’s all they can do, not least because they know remarkably little about either Gerry or Marco, beyond their respective job titles, companies, maybe a few qualification questions you asked on the registration form, and possibly some stuff it managed to scrape off LinkedIn without getting permission or paying LinkedIn any money (good luck with that).

They promise ‘deep compatibility’ matching users based on ‘nuanced interests’, but most of the time they just pair people with the same job title. Which, when you think about it, is entirely pointless.

A phrase I heard recently in a presentation for one of these tools is ‘serendipity at scale’.

Oh dear. We’ll gloss over the fact that the marketer who came up with this line doesn’t quite understand the definition of ‘serendipity’, but you get the point. They’re promising the magic of a chance encounter, but the mechanism that’s trying to bring this ‘serendipity’ about inherently lacks the emotional intelligence to understand intrinsic human motivation, social chemistry – or even just vibes.

The other claim made by these products – which I would very much dispute – is that of ‘effortless networking’. The lie here is in the suggestion that the AI does all the work and I magically happen to meet interesting people. But it doesn’t! And I don’t!

On the contrary, in my experience of using these things, matchmaking consists of little more than forcing users into exhausting ‘evaluative’ modes, trying to decide whether to accept or reject automated requests from an app to reach out to strangers who may (or may not) have any interest in meeting me.

For me, it’s roughly as welcome as the ten unsolicited ‘requests to connect’ that I get from total strangers on LinkedIn every week – all of whom either want to sell me some ill-defined lead-generation service, or a pension. Most of us would happily pay more not to be spammed with this junk outreach.

The problem is that it might be ‘effortless’ to send these requests to people but it’s kind of exhausting to receive them.

Of course, none of this should surprise us.

AI matchmaker apps are bound to struggle because they’re attempting to automate the complex, organic process of human networking with rigid algorithms. And while they promise ‘curated connections’ they mostly result in automated spam, missed opportunities, and privacy concerns that alienate attendees – especially the higher-value ones who (surprise surprise) represent the most obvious targets when it comes to nosiness networking.

This is why most people switch matchmaking off, and if they can’t switch it off? They stop using your app.

AI matchmaking under-delivers

Like most AI-centric SaaS, event matchmaking tools tend to be developed by young men in Patagonia gilets who still harbour some bitterness that they didn’t get into Stanford.

They can talk at some length about how the latest version of Claude has ‘moved away from vector DBs and is now stuffing the context window’ but they know roughly nothing about events – particularly the behaviour and goals of individual attendees at events.

Consequently, their products don’t do well when confronted with reality.

AI, while great at identifying patterns, really sucks at understanding context. So it’s not uncommon for a bot to schedule back-to-back meetings on opposite sides of a massive convention centre, or to try to book a high-stakes introductory meeting at 7am on the final day of a conference when the other person is either fed up with the whole thing or en route to the airport.

Then there’s the sheer spamminess of having an app repeatedly contact people to tell them who they should be meeting. It sounds like a good idea until you think about it for … like a minute.

For high-value attendees – the investors, CEOs, and industry leaders everyone wants to meet – the AI matchmaker is just a nightmare. Precisely because the app makes it ‘effortless’ to reach out, these guys are bombarded with generic, automated requests all day long.

When the barrier to entry for an introduction is lowered to a single swipe, you can’t be surprised that the quality of those introductions plummets! This leads to what our UX friends call ‘notification fatigue’, where the very people who make the conference valuable simply turn off the app or delete it entirely. Way to p*ss off those VIP stakeholders!

I know what you’re thinking. Maybe it’s a tad annoying for the rock star attendees everyone wants to meet – but it’ll be great for more junior attendees looking to get a start in the industry, right?

Wrong! For first-time attendees or those from emerging sectors, smaller NGOs etc, the AI has no historical data to work with. So these people often find themselves receiving really poor or plain random matches – or just no matches at all. The algorithm rewards those with established, keyword-rich profiles, effectively gatekeeping the community from the fresh perspectives it needs to thrive. This is what our UI friends call the ‘Cold Start’ problem and again, it’s really not what you want.

‘I know guys, let’s get some tech that will inundate our leading industry players with spammy sales pitches and treat everyone else like crap! And it’s only $20,000!’

Even if these were not problems (and honestly, they are) by constantly matching similar profiles – and always prioritising the purchaser / provider interface – AI networking tools end up reinforcing homogeneity while preventing the diverse networking that makes conferences valuable in the first place. You get an echo chamber effect. Look at Facebook and ask yourself if that’s the direction you want to take your event in.

Oh, and people don’t really trust it

Speaking of Facebook …

Many of these platforms seek to ‘enrich’ the profiles of your attendees by scraping data from LinkedIn, Facebook, and pretty much any other databases they can find, without explicit, granular consent for each data point used.

When an attendee sees a match based on a private interest or a former job they didn’t list in their event registration, the feeling is one of being surveilled rather than being guided towards a helpful networking opportunity.

We’re living in an era of heightened data sensitivity, and looking up a bunch of data about me to show to other attendees, when I didn’t ask you to, or even explicitly consent to it, is just creepy.

Also, a lot of this scraped data will inevitably be out of date or totally wrong. LinkedIn spends hundreds of millions of dollars a year preventing third-party, AI nonsense apps like these from scraping the profiles of its users. It’s incredibly difficult to mine this data accurately at scale. So what happens? They get the wrong Ian Dickie, or they read something I did in another industry 15 years ago, or … Garbage in, garbage out.

And while the tech bros may be really excited by the idea of their product being a ‘Black Box’ most of the rest of us are more sceptical. Matchmaking apps provide their ‘suggested for you’ lists without explaining the logic behind them.

When I get matched with a specific vendor, I do have to wonder if it’s because they truly fit my needs, or because they paid for a ‘Premium Exhibitor’ package that prioritises their visibility in the algorithm. When the financial incentives of the organiser and the vendor are hidden within a proprietary algorithm, the attendee quickly becomes more ‘product’ than ‘participant’.

OK, that’s enough negativity – what should I do instead?

If you really want to actively drive peer-to-peer networking at your event (but minus the AI spam slop) you can do what an increasing number of the top conferences do: that is deploy a human-in-the-loop matchmaking system instead.

This approach replaces the rigid algorithms with a ‘concierge’ style service where human experts who have a deep understand of the nuances of the business / technical / scientific needs of the attendees – facilitate higher-value connections.

Your event technology can help with the initial admin by suggesting a shortlist, but then a human expert takes over to ensure the connection makes sense in context and is likely to be mutually desired. Think of it as ‘curation over automation’. Listen to that rhyme! I’m like the Robert Frost of event tech blogs.

Other mechanisms include ‘curated marketplaces’. Never mind the terrible name, the idea here is simple. You have your app list a digital ‘catalogue’ of opportunities where participants manually browse and filter detailed business / research offers or needs posted by others. For example: ‘I am looking for a private sector research partner with expertise on laser absorption spectrometry to join an EU funding bid for …’

If someone is interested or can help in any way, they’re totally going to reach out and have coffee with me; and this is real networking with the potential to create real interactions and genuine value. It’s also very cheap and easy to achieve for you as the organiser.

And of course there are a ton of existing, tried and tested format ideas you can deploy like themed roundtables (facilitated small-group discussions led by a human moderator who ensures every participant has a chance to share their expertise); so-called ‘braindates’ (where attendees post topics they can teach, allowing others to book a 1:1 or group chat based on shared intellectual curiosity).

I’ve even been to a conference where the host association had what they called ‘Networking Ambassadors’ – volunteer hosts whose role was to proactively spot individuals standing alone during breaks and introduce them to relevant circles or exhibitors.

I’m not arguing that we should banish technology from the congress centre – that would be an interesting position for someone who runs an event tech company.

But I do think we need to work on app features that enhance rather than simply trying to replace human initiative.

AI is a fantastic tool for administrative tasks – the repetitive stuff that eats up time. We should totally be using it to manage room capacities, send reminders for sessions the user has already chosen, analyse data, and more.

When it comes to human connection though, we should be wary of trying to force ‘serendipity’ because it won’t work and it risks alienating customers.

The goal of a conference is to create a community, and a community is built on shared experience, and genuine interest. No algorithm, however advanced, can replace the spark of a real conversation.

So let’s stop trying to automate the chemistry and start focusing instead on building events where it’s more likely to happen naturally.