For Bullhorn users

Bullhorn AI without the per-message meter

Bullhorn's generative AI runs on a model you bring and pay for by the token. For a busy desk, that turns every draft, every pitch, every enrichment into a line item. Here's how that billing actually works — and the flat-cost way to run the same workflows.

Short answer: Bullhorn AI Assistant (formerly Copilot) uses a "bring your own model" design that routes generations through your own Azure OpenAI deployment, so the per-token cost of every message lands on your Azure bill. BraxtonCRM runs the same categories of AI work on local models on hardware you control — so heavy use doesn't run up a meter.

How Bullhorn's AI billing actually works

Bullhorn's AI capabilities — the generative drafting, screening questions, and candidate pitches inside the ATS — are delivered through Bullhorn AI Assistant (formerly Copilot), part of the broader Bullhorn Amplify lineup. The important detail for budgeting is the architecture: it's a bring-your-own-model implementation. Bullhorn doesn't host or train the model; it connects to an Azure OpenAI (Azure GPT) deployment that you own and control.

That design has genuine upsides — your data stays inside your own model instance, and you're on enterprise-grade infrastructure you don't have to build. But it also means the token cost of every generation is yours. Each draft, rewrite, summary, and pitch consumes tokens, and that usage is billed to your Azure account. The more your recruiters lean on it, the larger that line item grows.

What "per-message fees" really mean on a busy desk

Token costs look trivial per message and stop looking trivial at volume. A recruiter who drafts twenty candidate notes, a handful of pitches, and a batch of screening questions in a day is generating hundreds of calls a week. Multiply across a team and across a quarter, and the "pennies per message" framing turns into a real, variable, hard-to-forecast number — one that grows precisely when your team is most productive.

The structural issue isn't the price of any single call. It's that the meter is always running, so there's a quiet incentive to use AI less — exactly backwards from what you want when you're trying to get more leverage out of every recruiter.

The flat-cost alternative: local models

BraxtonCRM takes the other path. Instead of routing every generation to a metered cloud model, it runs open large language models locally, on hardware you control. The cost is the flat, known cost of running that machine — not a per-token charge that scales with how hard your team works.

Functionally, it covers the same ground recruiters actually use AI for: drafting and rewriting outreach, summarizing candidate history, generating screening questions, and powering candidate-to-job matching with custom taxonomies and geocoded distance scoring. The difference your finance team feels is the billing model — fixed instead of metered.

Side by side

 Bullhorn AI Assistant (Copilot)BraxtonCRM
Billing modelPer-token, via your Azure accountFlat cost (local hardware)
Cost when usage spikesRises with usageUnchanged
Where the model runsYour Azure OpenAI deploymentHardware you control
Works on Bullhorn dataYes (native)Yes (Bullhorn-native)
Candidate matchingSource & match AI (rolling out)Local-LLM matching + geocoded scoring
Infra to manageNone — Bullhorn-managed connectionRuns on your machine

This is an honest trade, not a free lunch. Bullhorn's managed path means there's no hardware to run and you're on first-party infrastructure. The local approach asks you to run a machine — in exchange for taking the meter off the table. For teams that use AI heavily, that trade tends to pay for itself.

You don't have to leave Bullhorn to try it

BraxtonCRM is Bullhorn-native — it reads and works with your existing Bullhorn data. You can run flat-cost AI workflows alongside Bullhorn rather than ripping anything out, and decide from real usage whether the economics make sense for your desk.

Frequently asked questions

Does Bullhorn charge per message for AI?

Bullhorn AI Assistant (formerly Copilot) connects to your own Azure OpenAI deployment under a bring-your-own-model design. Bullhorn doesn't host the model, so the per-token cost of each generation is billed to you through Azure. More usage means a larger token bill.

Is there a Bullhorn AI alternative without token fees?

Yes — BraxtonCRM runs comparable AI tasks on local models on hardware you control, so the cost is flat rather than metered per message.

Do I have to leave Bullhorn?

No. BraxtonCRM is Bullhorn-native and works on top of your existing data. You can run flat-cost AI alongside Bullhorn instead of replacing it on day one.

See the flat-cost version on your own data

A short walkthrough on your Bullhorn setup — no token meter, no pressure. See what your team's AI usage would actually cost.

Book a walkthrough