Help Desk Software's AI Arms Race, From Dixa to Oracle

Buying help desk software used to be a boring decision. You picked the tool with the tidiest inbox, wired up your email, and moved on. In 2026 the pitch has changed under everyone's feet. Every vendor on the shortlist now leads with the same promise: an AI agent that does not just route your tickets or suggest a reply, but closes the case on its own. The question a support leader asks has flipped from "which inbox do my agents like?" to "how much of this can I stop paying humans to do?"

That shift is why a lineup as strange as Dixa, Front, Thena, Oracle, InvGate and Jitbit ends up on the same page. On paper they barely belong in the same market. One is a Danish contact-center platform with nine figures of venture money. One is a shared inbox from San Francisco. One is four years old and lives inside Slack. One is Oracle. One is an ITIL service desk built in Buenos Aires, and one is a bootstrapped ticketing app from Edinburgh that you can still install on your own server. What lines them up is the bet they are all making at once - that the interaction, not the seat, is the thing worth automating.

The FieldSame market, six instincts

Start with the two that look most alike and aren't. Dixa, founded in Copenhagen in 2015, sells itself as conversational customer service - phone, chat, email and social pulled into one workspace, tuned for e-commerce and subscription brands that live and die by response time. It raised a $105M Series C led by General Atlantic and then went shopping, absorbing Elevio, Miuros and Solvemate to bolt intelligence and automation onto the core. Its 2025 answer to the AI moment is an agent named Mim that resolves inquiries by itself while keeping a person able to step in.

Front, born in San Francisco in 2013, took the opposite instinct. Instead of building a help desk that feels like a help desk, it kept the familiar shape of email and slid the workflow underneath - assignment, routing, collaboration, shared drafts. Founders Mathilde Collin and Laurent Perrin turned that into a business estimated around a $1.7B valuation, and in 2025 folded its automation into a single Front AI brand with an Autopilot agent the company says can resolve a large share of requests. The claim is a vendor figure, but the direction is unmistakable.

The interaction, not the seat, is the thing every one of these companies is now trying to automate.

Then there is Thena, the youngest name here and the one built on the sharpest observation. Its founders noticed that B2B customers had quietly stopped filing tickets and started pinging their vendors in Slack. Rather than fight that, Thena built the help desk around it - turning shared Slack and Microsoft Teams channels into a routed, SLA-tracked support system with AI triage on top. Backed by a $5M seed from Lightspeed and First Round, it is a bet that where the conversation happens matters more than the inbox it lands in.

The SpectrumFrom rent-it-all to own-it-yourself

At the far end of scale sits Oracle. Its Fusion Service product is not a standalone company but a slice of a suite, and that is exactly the point: service data wired directly into the same cloud as your ERP and the rest of Oracle CX. Oracle has moved hard on agentic AI, adding role-based agents like a triage assistant, opening a Fusion Applications AI Agent Marketplace with more than a hundred partner agents in late 2025, and announcing Fusion Agentic Applications for CX in April 2026. If your company already runs on Oracle, service is the next tile in the mosaic.

InvGate plays a different game again. It is an IT service management platform - incident, request, change, problem, knowledge and asset management, the full ITIL vocabulary - built in Buenos Aires since 2008. For years it was bootstrapped and profitable before taking a $35M growth round from Riverwood Capital in 2023. In 2025 it was named a Gartner Peer Insights Customers' Choice for ITSM Platforms with a 4.9 out of 5, and its Q4 release leaned into governed, controllable AI rather than a black box. It is the credible, well-run alternative to the enterprise incumbents.

And then Jitbit, which in this company reads almost like a provocation. Founded in Edinburgh in 2005 and bootstrapped ever since - no venture capital in twenty years - it sells a lightweight, email-based help desk you can run in the cloud or, crucially, install on your own hardware under a one-time perpetual license. In an all-cloud market, being one of the last real self-hosted options is not a limitation. For teams with data-residency or compliance constraints, it is the whole reason to call.

The Cheat SheetWho each one is really for

Strip away the marketing and the choice gets simpler. It is less about which tool has the most AI and more about which one matches how your customers already behave and how much of the stack you want to own.

*Prices are indicative, drawn from public vendor information as of mid-2026, and change often. Dixa and Thena are largely quote-based. Verify current figures with each vendor.

The Real ShiftWatch the pricing page, not the demo

The demos all look the same now: a customer asks a hard question, an AI agent answers, everyone nods. The more honest signal is on the pricing page. For two decades support software was sold by the seat - more agents, more money. The AI wave breaks that model, because the entire pitch is that you need fewer agents. So the industry is drifting toward outcome pricing, charging per resolved conversation rather than per human logged in. When a vendor tells you their agent resolves most tickets and still wants to bill you per seat, the incentives are quietly working against you.

When the software's whole promise is fewer agents, per-seat pricing is the tell that the incentives don't line up.

The other thing worth holding onto is skepticism about the numbers. A resolution rate of "up to 70%" is a marketing ceiling, not a guarantee, and how close you get depends heavily on your own data, your documentation and the messiness of your customers' questions. InvGate's framing of controllable, auditable AI is a reaction to exactly this - buyers want to see the machine's work, not just trust it. The vendors racing hardest on autonomy will also, in the end, be judged on how gracefully they hand a conversation back to a person when the agent is out of its depth.

So the field of six is really a set of questions pointed back at you. Do your customers email, call, or live in Slack? Do you want service fused into an enterprise suite or a light tool you can run yourself? Do you trust an AI agent to close a ticket unsupervised, and would you know if it got one wrong? Answer those, and the shortlist picks itself. The arms race is loud, but the winning move is boring - buy the tool whose defaults already match how your customers behave.