AGENT FIRST
How It Works Industries Free Tools Agent Challenge Deal Room Pricing Apply for the pilot program
Pilot program

Controlled evaluation first.
Production data only after the security gate.

Agent First pilots begin by observing a bounded slice of how work actually happens before anything is configured. We map the decisions, searches, handoffs, exceptions, and governing knowledge; establish the baseline; then build and test Agent First around that reality. Evaluation can start with fictional or de-identified material. Proprietary production client information waits for Security Readiness sign-off.

This is a paid operational pilot, not a free trial: you use Agent First at the standard seat rate while we measure it together on your real rules and real calls.

Up to 90 days$69 per pilot agent per month10-seat minimum ($690/month)$2,500 Profile fee waived$500 conversion credit
Apply for the pilot program Existing pilot partner? Sign in
The pilot

Use the smallest real workflow that can produce useful evidence.

1 workflow
One bounded workflow family and one primary Client Profile by default. Additional Profiles or integrations are separately scoped.
Up to 90 days
The live production pilot clock starts only after operational readiness and, where proprietary production data is involved, Security Readiness Gate sign-off.
$69
Per pilot agent per month at the Starter rate, with a 10-seat minimum ($690 per month). No Growth-tier discount during the pilot. No automatic renewal.
Security diligence available now

Buyers can review the current state before the production gate.

Agent First maintains a current-state Security, Data, and Governance package covering system and data flow, data inventory, preliminary vendors, role and client separation, rule governance, security-questionnaire responses, and the production verification backlog. It is shared through the private Deal Room.

Open the Security, Data, and Governance package (Deal Room sign-in)
Security Readiness Gate

Before production data enters Agent First, the environment must earn it.

Agent First can be demonstrated, configured, benchmarked, and tested with fictional or appropriately de-identified material. Before proprietary production procedures, client-specific operating material, sensitive production context, or regulated information enters a live pilot, Agent First completes a focused security architecture and application-security review with an experienced security-minded SaaS engineer or appsec consultant.

  • Threat model and architecture/data-flow map.
  • Server-side Agent, Supervisor, and Admin authorization, plus company/tenant and Client Profile isolation testing.
  • Review of authentication, sessions, APIs and database, secrets, uploads, AI/model-provider data flow, logs, backups, retention and deletion, dependencies, and incident response.
  • OWASP ASVS 5.0 mapping, critical and high finding remediation, retest, and a signed data boundary.
Not SOC 2 tomorrow: the sequence is prototype and demo, then security architecture review, then critical remediation, then a hardened pilot environment, then real production pilots, then formal compliance when buyers actually require it.
What happens

Nine steps from interest to operating decision.

Step 1
Scope the observation
Choose the bounded workflow, sponsor, representative users, permissions, and the question the pilot needs to answer.
Step 2
Observe real work
Watch a representative slice of the workflow before configuration. With approval, workflow-capture tooling can document systems, searches, handoffs, delays, workarounds, and decision points.
Step 3
Map decisions and sources
Connect each material decision to required context, exceptions, authority, approved sources, next actions, and escalation. Resolve differences between documented, observed, and required work.
Step 4
Lock the baseline
Measure the pre-intervention workflow using an agreed method before Agent First results are visible.
Step 5
Build the Client Profile
Bring what you have, as it is: scripts, policies, transfer rules, and contracts as PDF, text, or spreadsheet files, or pasted in. No reformatting. Agent First drafts the Client Profile around the observed decision moments, flags anything unclear for review, and uses approved, fictional, or de-identified material where appropriate.
Step 6
Test calls, sign-off, and Security Gate
Your supervisor runs about 25 real scenarios through Agent First: the common questions, the ones people get wrong, and every must-not-miss rule. Anything that fails is fixed and retested, and nothing goes live until they sign off. The same scenarios run again whenever a rule changes. Required security and data gates pass before proprietary production use.
Step 7
Live pilot
Bounded use, weekly review, issue tracking, rule governance, and measurement.
Step 8
Final readout
Compare the baseline with results, limitations, issues, and what the workflow taught us.
Step 9
Decision
Continue, expand, revise and retest, or stop.
What gets measured

Five agent measures, plus how the system performs.

The baseline locked in step 4 and the readout in step 8 use the same measures in every Agent First pilot, not a different scorecard per customer. Success thresholds are written down before any results are visible.

Decision resolution time
How long it takes to go from a question arriving to the correct, client-specific answer being given. Timestamped at the decision moment, before and after configuration.
Correct-client rule application
Whether the rule actually applied matches the rule that governs that specific client, account, or location. Sampled and checked against the source policy, not self-reported.
First-contact accuracy
Whether the first answer given was the right one, versus one that had to be corrected, escalated, or redone on a follow-up contact.
Supervisor pull-ins
How often a frontline worker needed a supervisor or senior teammate to resolve something a correct, available answer could have handled directly.
New-hire ramp time
How long it takes a new or reassigned worker to reach full independent accuracy on the piloted workflow, compared to the pre-pilot baseline.
How the system performs
Time to go live (days from your material to supervisor sign-off), adoption (share of pilot agents using it each week), answer speed (seconds from a lookup to the answer on screen), and test-scenario accuracy at go-live and after every rule change.
Your customer results
If you already track satisfaction, complaints, or callbacks for the piloted workflow, they're reported alongside, as observed during the pilot, not claimed as caused by Agent First.
How it's collected
A rotating sample of sessions each week, not continuous monitoring of every worker: live shadowing, recorded-interaction review, supervisor pull-in tallies, and test-scenario results, cross-checked against each other.
Commercial terms

What you sign up for, and what happens after.

Pilot
  • Up to 90 days live after operational readiness and, where proprietary production information is involved, Security Readiness Gate sign-off.
  • $69 per pilot agent per month, 10-seat minimum ($690 per month).
  • $2,500 Client Profile implementation fee waived.
  • No Growth discount during the pilot.
  • One bounded workflow family and one primary Client Profile by default; additional Profiles, major integrations, or expanded scope are separately scoped.
After the pilot
  • $500 credit on the first post-pilot invoice.
  • Validated Profile carries forward.
  • Starter $69, qualifying Growth $49, Enterprise custom.
  • Tier is assigned based on deployment scope, not self-selected.
  • No automatic renewal.
Who fits
Rules-heavy, source-ready, measurable, sponsored, and bounded operations where the correct answer changes by client, location, account, product, service, or exception.
Accepted pilot partner?
Use the same private account from the Deal Room and demo to complete onboarding, lock the baseline, record weekly measurements, and finish the readout.
Open the private Pilot System → Apply for the pilot program
AGENT FIRST
Keep the human. Upgrade the agent.
Agent First Consulting →