Dixa’s journey: a Q&A with Christian Lohmann

Christian Lohmann is CEO of Dixa, the agentic customer service platform used by 1,000+ ecommerce brands across 42 countries. Before joining Dixa, he spent nearly a decade in M&A and corporate finance at Carnegie Investment Bank, TDC Group, and Danske Bank.

Tell us about Dixa and what problem you’re actually solving

Many ecommerce brands have a serious customer service problem they don’t fully admit to themselves. They say they care about the experience – and I believe they believe that – but then you look at their refund policy, how hard it is to actually contact them, the phone number buried at the bottom of the page. And then you look at their Trustpilot score and you wonder where the gap is.

Dixa is an agentic customer service platform built for ecommerce. The mission is to build long-lasting relationships between brands, their customers, and the people doing the work on the inside. In practice that means omnichannel conversations, AI that can handle queries end to end, a unified workspace that gives agents full customer context, and the analytics to continuously improve. Over 1,000 brands across 42 countries use it.

What drew me to it: I came from M&A and corporate finance, so I spent a lot of time looking at what’s actually going on inside businesses versus what the board presentations say. That gap between what companies claim about their customer experience and what their Trustpilot score says has always interested me. This felt like the right place to do something about it.

What’s the thing about AI in customer service you wish people would stop saying?

That it’s the low-hanging fruit. Everyone says it: high volume, repetitive queries, easy to automate. I think it gets the complexity of the role almost exactly backwards.

Here’s another way to think about this: a CFO is often following well established rules. Rules are trivial to automate – you read the document, you check if something is inside or outside the boundary, you apply the policy. A customer service rep is making human judgements in ambiguous, emotionally charged situations, often on behalf of a brand that hasn’t given them the authority to actually fix the problem. That is far harder to automate well.

The work AI should eliminate is the stuff that never needed a human in the first place: order tracking, status updates or return initiation. Brands should have automated that five years ago without any AI. The mistake is assuming that because you can deflect 80% of queries, you should – and that doing so constitutes a good customer experience.

You’re selling AI but you’re also warning people about its limitations. Don’t those two things contradict each other?

I think you have to be honest about it. The quality and accuracy improvements from AI are real. But the cost case – “AI is dramatically cheaper than people” – is being built on a pricing assumption that may not hold. I believe that token costs right now are subsidised by investor capital. The major LLM providers are not pricing at sustainable margins. If you’re building a board presentation around today’s API costs as a permanent baseline, you’re taking on hidden risk.

The brands doing AI in customer service well understood that they had to redesign what they were willing to do for customers and then used technology to deliver it at scale. The technology is secondary to the decision about who you want to be as a brand.

So what actually goes wrong when brands integrate AI in their customer service without fixing the underlying issues first?

If a brand has a policy that customers hate like a restrictive refund process, a complaints procedure designed to make people give up – automating the delivery of that policy doesn’t fix it. The customer is still not getting their money back. They’re just being told so more efficiently.

The support team isn’t usually the problem. They’re often smart, caring people. But they’ve been set up to deliver a policy, not a good experience. No amount of AI fixes that. You often get the Trustpilot score you deserve.

What’s been the hardest part of building the business?

When I joined, I had to spend a couple of years doing what I’d call the Excel work: operational efficiency and reducing cash burn. It was hard and un-glamourous but I think the business needed it.

The truth is that lots of companies in our category haven’t made it. The CX software market looked very different a few years ago, and surviving the consolidation while actually building something better has been the real challenge. AI has helped us – not because it made us cheaper overnight, but because it genuinely changes what the product can do for customers.

What does the relationship between a brand and its CX vendor look like today?

Modern software isn’t a product you buy and drive away in. It’s a starting point that requires daily optimisation and a vendor who stays in the room after go-live. The gap between what a CX platform can do and what your team actually does with it is almost always a vendor problem – but sometimes it’s also that the buyer doesn’t actually want a strategic partner. They want a login and someone to call when it breaks.

See Also

AI breakthrough turns smartphones into lifesavers for pets

What’s interesting is that we have the same problem ourselves. We buy a lot of software for our size and honestly we treat a lot of it transactionally – don’t engage deeply, then complain when we’re not getting the value we expected. Which is exactly what some of our customers do with us. That loop is something I think about a lot. It’s a relationship problem rather than a technological one, and it goes both ways.

What’s your single biggest disagreement with how most companies are approaching AI right now?

The obsession with deflection rate as the primary metric. All it measures is whether you stopped the customer from reaching a human – not whether they got what they came for, not whether they’ll come back. Optimising for that number while ignoring outcome quality is how you automate yourself into a bad reputation.

My broader point is: treat AI adoption the way you’d onboard a new employee. You don’t hand someone their hardest cases on day one. You build capability gradually, check the quality of what they’re doing, expand their responsibilities as trust is earned. Most brands are doing the opposite.

What are you most interested in outside of the day job?

The speed of innovation under pressure. I’m involved in a defence innovation initiative working with Ukraine on accelerating tech development – what happens when the normal product cycle doesn’t exist, when the feedback loop is weeks not years, when the stakes are existential. It changes how you think about building things and making decisions fast.

What would you go back and tell yourself at the start of your career in this space?

That the culture of how you treat customers starts with whether the CEO thinks the support team matters. In too many companies, they don’t. Every policy decision, every budget cut, every response time target reflects that. The technology will not save you from that. And I say this as someone who sells the technology.