Surfaced signalsdepth · confidence
Renewal risk
Pricing delta
Context & history
Surface chatter
from 25 sources3 decisions surfaced
Operational AI infrastructure
Agents that learn your operations.
Vesper reads every document, thread and ticket your team touches, then runs the work end to end — and hands you the decision, not another digest.
No card required · Live in an afternoon · Cancel anytime
Vesperreading 3 sources
The renewal terms changed in v4. Auto-renewal is now opt-out, not opt-in — this materially shifts risk. Worth a legal flag?
Flag it, and what else moved?
Ask across all sources…
ReadingThreads 13 · Docs 8
Master v4Auto-renewal clause redlined — §7.2
#deal-acmeLegal asked about liability cap on Tues
Q3 board“We should close Acme before quarter”
Redline14 changes since v3, 3 flagged material
Last sweep 40s agoAuto
4.2M+workflow runs completed
92%less manual triage per team
180+operations teams onboarded
Benefits
Not a chatbot. A colleague with a job description.
Every Vesper agent gets a scope, a set of sources and a definition of done. What it produces is a decision with its working shown — never a wall of text you still have to read.
Learns from your work
Point it at a folder, an inbox or a channel. It builds its picture from what you already have — no schema, no data project, no six-week onboarding.
Runs on a schedule
Hourly sweeps, nightly reconciliations, a check every time a document changes. Agents work while nobody is watching and report only when it matters.
Shows its working
Every claim links back to the line it came from. Reviewers audit the trail instead of taking the model's word for it — which is the only way this survives compliance.
Escalates, doesn't guess
Confidence below your threshold, or a decision above your value limit, and the run stops with a question for a human. Silence is never mistaken for approval.
Your data stays yours
EU or UK residency, no training on your content, retention you set in days. Delete a workspace and the derived indexes go with it inside 24 hours.
Fast to first value
Most teams have an agent doing real work before the end of the first afternoon. The slow part is deciding what to hand over, not the setup.
How it works
Three steps to a working agent
Scope it, connect it, let it run. You keep the judgement calls; it keeps the legwork.
STEP 01
Describe the job
In plain language: what to watch, what counts as important, what to do about it. Vesper turns that into a scoped run plan you can edit line by line.
STEP 02
Connect the sources
Drive, mail, chat, tickets, your warehouse. Read-only by default, scoped to the folders you name, revocable from one screen.
STEP 03
Let it run, review the trail
Watch the first few runs, approve or correct, then raise the automation level. Corrections carry forward — that is the whole learning loop.
Live demo
Watch an agent work
Three agents from a real deployment shape. Switch between them — the run, the metrics and the confidence all recompute.
Run trace
- 01Pulled 25 sources: 8 documents, 13 threads, 4 calls0.8s
- 02Diffed master agreement v3 → v4, 14 changes2.1s
- 03Classified 3 changes as materially adverse3.4s
- 04Matched §7.2 against the approved fallback list4.0s
- 05Drafted the redline note and flagged legal5.2s
Escalated to a human: auto-renewal moved from opt-in to opt-out, which sits outside the pre-approved fallback positions.
Handled without a human96.4%
Median run time5.2s
Hours returned per week31
Demo data. Figures are illustrative of a mid-market deployment, not a live account.
Pricing
Priced per agent, not per seat
Invite the whole team from day one. You pay for the work being done, not the people watching it.
Pilot
£0/mo
- 1 agent, 200 runs a month
- 3 connected sources
- 7-day run history
- Full audit trail
- Community support
Team
£240/mo
- 5 agents, 10,000 runs a month
- Unlimited sources and seats
- Scheduled and event-driven runs
- Escalation rules and approvals
- Priority support
Scale
£880/mo
- Unlimited agents and runs
- SSO, SAML and SCIM
- EU or UK data residency
- Custom retention and export
- Named solutions engineer
Prices in GBP, excluding VAT. Run overages are never billed silently — we ask first, every time.
FAQs
Good to know
What is a "run"?
One complete pass of an agent over its scope — read the sources, do the work, produce the output. A nightly contract sweep is one run a day; an agent triggered by every inbound ticket is one run per ticket.
Do you train models on our data?
No. Your content is used to answer your questions and nothing else. Corrections improve your workspace's own instructions, which stay inside your workspace.
What happens when the agent is unsure?
It stops. Every agent has a confidence floor and a value ceiling you set. Below or above them the run pauses with a specific question and a link to the evidence it was weighing.
How long does setup actually take?
Connecting sources takes minutes. Deciding what you are comfortable handing over takes longer — most teams pilot one narrow job for a fortnight before widening the scope.
Can we self-host?
Not today. Scale customers get single-tenant infrastructure in the region of their choosing, which covers most of what teams actually mean when they ask.
Hand over the first job
Pick one workflow nobody enjoys. Have it running before the end of the day.
No card required · 14-day Team trial · Cancel anytime