Toma IQ: The AI that actually learns how your dealership operates

Every AI vendor says their system "gets better over time." Here’s what that actually means, and why Toma IQ is different.

Dealers are seeing something surprising in the data: voice AI can outperform humans on service calls until edge cases arise.

Pied Piper's 2025 Service Telephone Effectiveness Study analyzed more than 2,100 dealerships and found that when AI handles a service call end-to-end, it scores 72 on their effectiveness index — 8 points above the national dealer group average. But when that same AI can't fully resolve the call and transfers the customer to a human, the score drops to 50 — 14 points below the dealer group average, and 22 points below what AI achieves on its own.

The AI isn't the weak link. The edge case handling is.

This is more a training problem, than a technology problem. For most dealership AI, it doesn't know when to escalate, who to route to, or how to keep the customer engaged while it sorts itself out. Nobody taught it how that specific store actually runs.

And here's the part that should bother any dealer running voice AI: the same mishandled edge cases happen again next week. And the week after. Because nothing about that failure taught the AI anything.

That's the problem Toma IQ solves.

Why "AI that gets better over time" is only half the story

When a vendor tells you their AI improves over time, they're usually telling the truth, just not the whole truth…

Most AI vendors rely on global model updates from the foundational model companies, which are real and powerful. When a new AI model is released every AI product does get incremently better. They hallucinate less, handle more conversational variety, and improve at intent recognition. All of that matters.

But a global update doesn't know that customers calling about a late-model F-150 recall need to be scheduled at least 2 weeks out. It doesn't know that a caller who mentions a fleet account should skip the standard scheduler and go straight to your commercial team.

Those things can only come from your calls, the actual workflows your team has developed, the edge cases your managers have resolved, the preferences your service director has built into the way your store runs. And until now, the only way to teach an AI about them was frustrating by design: sharing multi-page training documents during onboarding that the AI can rarely apply reliably in real conversations, manually flag calls and submit support tickets, and wait days — sometimes weeks — to see if anything changed. While you wait for the AI vendor to update your agent, the same scenarios kept surfacing on live calls.

The real gap in dealership AI isn't the underlying model. It's the learning loop.

Introducing Toma IQ

Toma IQ is an intelligence layer built on reinforcement learning with human feedback (RLHF), the same technique that powers the most capable AI systems in the world, now applied specifically to how dealerships operate.

The core idea behind RLHF is straightforward: instead of improving through generic training data, an AI agent learns from actual outcomes in its specific environment, guided by human judgment. When something goes well, the system learns to repeat it. When a human intervenes and handles a situation differently, the system learns why that was the better call. The AI gets smarter by doing, not just by receiving updates from afar.

By 2025, 70% of enterprises had adopted RLHF or similar alignment techniques to improve AI performance, up from just 25% in 2023. It's how OpenAI trains GPT, how Anthropic trains Claude, and how the best customer service AI agents across other industries learn to behave the way customers and teams actually want. Toma's founding team built their careers on this technology at Scale AI, Uber, and Amazon before applying it to automotive.

Toma IQ is the first time RLHF has been applied to the specific workflows, policies, and edge cases of automotive retail.

How Toma IQ works in practice

Toma IQ does three things continuously in the background.

It analyzes your calls. Every call your dealership handles gets reviewed for patterns: scenarios where a human intervened in a specific way, edge cases that recur across multiple calls, situations where a particular response consistently produced a better outcome.

It generates plain-language policy recommendations. When a pattern worth codifying appears, Toma IQ drafts a specific, actionable recommendation your AI Coworker can adopt. Not a training document. Not a configuration file. A plain-English policy your service manager can read and either approve or reject.

A real recommendation might look like this: "When a customer asks about loaner vehicle availability and mentions needing it for more than five days, offer to check with the service manager before confirming."

It puts you in full control. Every recommendation lands in your My Policies hub, a central library inside your Toma account settings page, and nothing changes until a manager approves it. One click to accept. One click to reject. Every change logged, every policy reversible at any time.

The whole process takes less time than reviewing a single call.

The difference between global intelligence and yours

The intelligence Toma IQ builds belongs to your store, not to Toma's platform.

When competing vendors release an update, every customer receives the same improvement. Their model configuration gets a little better at general conversation. That's real progress. But it means your AI agent is exactly as smart about your dealership as every competitor's AI agent is about theirs.

Toma IQ doesn't work that way. Every policy recommendation is generated from your calls, reflects your team's handling preferences, and requires your explicit approval before anything changes. A policy approved at your Honda store doesn't affect your Chevrolet store next door unless you apply it there. The intelligence your store builds over time belongs to your store.

The best AI Coworker for your store is the one trained on your store, not a global update that treats your dealership the same as every other dealership in the country.

What that creates is compounding advantage. The longer you run Toma, the more your AI Coworker has been shaped by how your operations actually work. That gap between your AI and a generic system widens every week and it can't be copied, because it's built on your data, trained on your call patterns, and approved by your managers.

What the My Policies hub looks like

Toma IQ surfaces in a dedicated section of your Toma account called My Policies. It's where your entire policy library lives: every approved policy your AI Coworker currently operates under, organized by topic: Loaner Policy, After-Hours Handling, Recall Escalation, Transfer Rules, etc. with full version history and the ability to revert any change at any time.

My Policies also supports manually written policies — if your team wants to define behavior proactively, you can write a policy directly and apply it to your AI Coworker.

You're not constantly creating policies, just reviewing specific recommendations the system has already drafted based on patterns it detected in your actual calls.

For General Managers and Fixed Ops Directors who want to know what the AI actually knows: My Policies is that answer. Every policy is listed. Every change is logged. The system is fully auditable.

For Service Managers and Service Advisors frustrated by edge cases that keep slipping through: this is how you close those gaps, permanently, without filing a support ticket or waiting on a vendor update.

Included in base pricing. Available today.

Toma IQ is included in the base pricing for Service Agent and Full-Store Receptionist. It's not a tier upgrade. It's not an add-on. It's how Toma works now.

Admins can access it through My Policies in Team Settings. If you've been running Toma for any length of time, there are likely policy recommendations already waiting for your review — Toma IQ has been analyzing your call history since launch.

If you're not yet a Toma customer and you want to see what AI trained on your dealership looks like in practice, we'd like to show you.

Request a walkthrough →

Frequently asked questions

Does Toma IQ create more work for my managers?

No. The AI does the analysis and drafts the recommendations. Your manager's job is to review and approve or reject — usually in less time than reviewing a single call. The goal is to reduce the number of times your team has to intervene, not add another task to their plate.

How do I know the policy recommendations are correct?

Recommendations are generated from patterns in your actual resolved calls — specifically, scenarios where human handling consistently produced better outcomes. Every recommendation requires your explicit approval. Nothing changes without your consent, and you can revert any policy at any time.

What if the AI learns something wrong?

It can't act on it without manager approval. The review step is the safeguard. Every policy is visible, logged, and reversible. If something slips through that shouldn't have, you can remove it immediately from My Policies.

Other AI vendors say their system improves too. What's different?

Most vendors mean the underlying language model received a global update. That improves general conversation quality across all their customers equally. Toma IQ learns your specific loaner policy, your after-hours transfer rules, your recall handling preferences — none of which a global model update can produce. The distinction is dealership-specific learning versus platform-wide improvement.

Is this included or extra?

Included. No add-on fee, no upgraded tier required. If you're a current Service Agent or Full-Store Receptionist customer, you already have it.