AI for energy origination vs procurement: two very different workflows
Origination teams hunt. Procurement teams negotiate. Both work the phones, both run email cadences, both live in the CRM — but the conversations, the data they care about, and the KPIs are completely different.
Trying to run both off the same AI agent is a mistake we see weekly. The origination agent ends up too transactional for procurement's relationship work; the procurement agent ends up too cautious to ever cold-open a new account. Neither side trusts the output.
What origination AI is actually for
Origination is the discovery side of the desk. The job is to find accounts that look like your last 50 closed-won deals, enrich them with supplier mix and renewal timing, and run cold outreach that earns a 5-minute discovery call.
An origination agent needs to be good at:
- Searching a target market with structured filters (region, load size, SIC code, supplier).
- Enriching companies with contract-end-date signals, decision-maker mapping, and public PR moments.
- Writing cold opener emails that prove the account was read, not templated.
- Cold-calling with a discovery-first script that books, not closes.
- Logging every touch against a contact so the next touch is informed.
Its KPIs are dial volume, meetings booked, account-creation rate, and reply rate on cold outbound.
What procurement AI is actually for
Procurement is the live-deal side. The job is to manage supplier matrix updates, chase quote returns, qualify inbound RFPs against your pricing rules, and keep every stakeholder in the thread without dropping balls.
A procurement agent needs to be good at:
- Reading inbound RFPs and structuring them into a quote-ready record.
- Following up with suppliers on outstanding quotes, with the right tone for an ongoing relationship.
- Calling existing customers about renewals 90/60/30 days out.
- Spotting matrix updates and flagging deals that need to be re-quoted.
- Keeping a clean audit trail of every conversation for compliance.
Its KPIs are quote-to-return time, supplier response rate, won-RFP percentage, and renewal contact rate inside the 90-day window.
Why one prompt can't do both
The two workflows pull on the model in opposite directions. Origination wants energy, persistence, and a willingness to open with a stranger. Procurement wants restraint, context-awareness, and the ability to manage long-running threads without losing tone.
An agent prompted to do both well almost always ends up doing both badly. The cold-outbound opener leaks into the supplier follow-up ("Quick question — are you the right person for this?" sent to a supplier you've worked with for five years), or the careful procurement tone neuters the cold outbound ("I don't want to take up your time, but…" sent to a stranger).
How to structure the stack
Deploy them as two named agents. Share the contact graph so a procurement contact who later becomes an origination target is recognized. Share the call history so the agent that calls a customer knows what was said last week. But keep the prompts, the KPIs, and the dashboards distinct.
- Origination agent owns the cold outbound queue, the discovery script, and the meetings-booked report.
- Procurement agent owns the inbound RFP queue, the supplier follow-up cadence, and the renewal pipeline.
- A shared inbound qualifier answers the main line, identifies the caller, and routes — to the human desk if the deal is hot, to the right specialist agent if it isn't.
The org chart matters more than the tech
Most teams that struggle with AI on the desk don't have a tooling problem — they have an ownership problem. If no one owns "the origination agent" the way they would own a hire, no one will tune its prompt, review its calls, or push back on its mistakes.
Treat each agent as a teammate with a job description. Review its work weekly. Iterate on its prompt the same way you'd coach a new SDR. That is what separates the desks where AI is producing revenue from the desks where it is producing demos.
