AI workflows
A workforce of AI agents, on your side of the table. Agents that negotiate your renewals, ghost write your vendor email, watch pricing and licensing changes across your vendors, and keep your paper honest. Twenty four standing agents work from day one, and your team composes its own agentic workflows without code.
Four departments, one rule: every agent shows its work.
Every run is logged, every figure is grounded in your stored data, and nothing leaves the platform without a gate you chose.
Your renewals, worked daily. Your replies, already drafted.
Hand a renewal to Vera in the Deal Agent. She works your renewal calendar every day at the autonomy you choose: watch and remind, draft for your approval, or negotiate by email inside the limits of a mandate you approve. She stops at every gate for your decision and brings every offer back.
Forward a vendor email and get back a reply worth sending, in a seasoned negotiator's voice, thread memory included. Grounded in your position: your benchmark standing, your terms, your leverage. Every draft waits in Mail, your own mailbox; you press Send, never Vera.
Eleven watchdogs on duty: pricing, licensing changes, billing, vendor health.
They run on the platform's clock, read only the data named on their card, and land their findings on the page they concern. You do not build them; you tune what each one tells you.
Analysts and ghost writers, on demand and on a schedule.
Six playbook specialists write the documents a deal needs, each in the executive brief standard, every figure grounded in your stored data. Beyond them, define up to six custom analysts: your prompt, your depth, your choice of which documents each may read.
An agent needs somewhere to work, and something it is allowed to do.
Autonomy without a boundary is a liability, so the agents work inside rooms with a stated goal and under a mandate you sign. Every draft is checked against that mandate before it can leave, and anything past the line stops and waits for a person.
One contract, one case
Press Create on any contract and its case holds the terms, the benchmark, the timeline, the strategy, the talking points, the research and the people, with every app opening on that contract.
What the machine did overnight
Renewal pressure with the notice brackets marked, every open thread sorted by who it waits on, the run feed, and the authority each agent holds. What the guardian stopped is at the top, with its reason.
Target, disclosure, autonomy, an annual authority cap, and an expiry date. Signed once per agreement, kept on the record even after it is revoked.
Reads every draft before it leaves. Allow, rewrite, block, or escalate, with the rule it applied and the reason in plain English.
Every action, who authorized it, what it committed, and where it went. The audit log is the same one your security team reviews.
Compose your own. Build the automation once, it works every week after.
Describe the goal in one sentence and the AI drafts the chain: trigger, steps, and the approval gate where anything would leave the platform. Test it against your own data before it earns a trigger.
Name it, pick a trigger, stack the steps.
Describe what the analyst step should cover in plain language, paste your own output template if you want the memo in your house style, and add steps with one click. Pause, edit, or run any workflow at any time.
Agents you can reach from any inbox.
Employees email a question to a private agent address; a white-paper answer comes back. Proposal price checks, discount sanity checks, risky-terms scans, should-we-buy advice, and more.
A named analyst that stays on the case.
Vera runs a grounded streaming chat over your contracts and the market, every figure cited back to stored data. Save an ask and turn it into a watch that re-runs on a schedule, so the answer lands before you think to ask again. Spoken mode reads the briefing aloud.
Every analysis ends with the emails already drafted.
Automation is not just the scheduled job. Wherever a result lands, a What next desk follows with the moves that fit it, built from that page's own numbers, so the step after the verdict takes minutes instead of another meeting.
The benchmark case in the Benchmark Library drafts six emails on the case's own numbers: the vendor, your CFO, procurement, your reseller, and your team, and the vendor facing drafts never reveal your rank or your sources. Sourcing writes the kickoff, the supplier invitations, the Q&A broadcast that answers every bidder equally, the chase with the quiet suppliers' names already in it, the recommendation memo, and the award and decline notes. Every draft waits in Mail, your own mailbox; you press Send, never Vera.
The Send it on drafter turns any report into the email that forwards it, written from the report itself. In the Document Agent, the workroom under every answer drafts the vendor email without showing your hand, briefs your CFO, invites colleagues to a working session with the agenda drafted from the cited clauses, and turns the finding into a 6 to 12 week project plan on a Gantt board, saved with the finding.
An open protocol for agent-to-agent negotiation.
ANP v0.1 is a published spec at app.vendorbenchmark.com/agent-protocol with an open API at /api/agent/v1, so your agent and a counterparty's agent can negotiate directly. Every exchange is written to a hash-chain ledger, an append-only record either side can verify.
Read the protocol spec →Your AI workforce is already hired.
Twenty four standing agents watch your portfolio from day one. Add your own.
The AI agent roster
Fifteen prebuilt agents, one page each.
What every agent does, the documents it writes, and the time it gives back, from the renewal war room to the weekly market monitor.
