Short answer: Bullhorn AI Assistant (formerly Copilot) is requested through your Bullhorn Account Manager and connects to your own Azure OpenAI deployment under a bring-your-own-model design. You grant user access through a permission set. The real work, though, is everything before that — clean data and a plan for cost and usage.
What the setup actually involves
The mechanics are more straightforward than the readiness. At a high level:
- Request the feature. Bullhorn AI Assistant (formerly Copilot) is enabled by reaching out to your Bullhorn Account Manager. It has been made available to Bullhorn Recruitment Cloud customers at no extra license cost.
- Connect a model. It uses a bring-your-own-model design and integrates with Azure OpenAI (Azure GPT). You'll need an Azure deployment for it to talk to — Bullhorn doesn't host the model.
- Grant access by permission set. Access is controlled through a permission set, so you decide which users or groups can use AI features rather than turning it loose org-wide.
Always confirm the current steps in Bullhorn's own documentation or with your account manager — product details change. This is the shape of it, not a substitute for their setup guide.
The part most teams skip: readiness
Switching it on is easy. Getting value from it depends on three things that have nothing to do with the toggle:
1. Your data has to be ready
AI reads your existing fields. If statuses are inconsistent, candidates are duplicated, or discipline and source fields are blank, the output inherits that mess. Run a cleanup pass first — we wrote a full Bullhorn data cleanup checklist for exactly this.
2. You need a cost plan
Because it runs on your own Azure model, every generation consumes tokens billed to your Azure account. Before you roll it out widely, decide who gets access and roughly what usage looks like, so the bill doesn't surprise you. If a metered model is the wrong fit for how heavily your team would use AI, a flat-cost local-model approach is worth comparing.
3. Your team needs a few good habits
AI drafting is only as good as what people ask for. A short internal cheat-sheet of prompts that work for your desk — outreach openers, screening questions, summarization — gets a team productive far faster than letting everyone rediscover it solo.
A quick readiness checklist
- Duplicate candidates and contacts merged
- Status picklists standardized and remapped
- Discipline / specialty fields normalized to a controlled vocabulary
- Azure OpenAI deployment provisioned and connected
- Permission-set plan for who gets access
- A rough cost expectation for your usage level
- A starter prompt cheat-sheet for recruiters
Frequently asked questions
How do you turn on Bullhorn AI Assistant (Copilot)?
Request it through your Bullhorn Account Manager. It connects to your own Azure OpenAI deployment (bring-your-own-model), and you grant user access through a permission set.
What do you need before enabling it?
An Azure OpenAI deployment, the feature requested via your account manager, a permission-set access plan, and — most importantly — reasonably clean data, since the AI reads your existing fields.
Is Bullhorn Copilot free?
It's been offered to Bullhorn Recruitment Cloud customers at no extra license cost, but because it runs on your own Azure model, you pay the per-token usage of every generation through Azure.