It Builds the Software It’s Missing
Theo notices Finance exporting from NetSuite to Excel to Stripe and back every month. It writes an internal service to replace the routine, tests it on past data and asks before deploying.
AI Employees That Do What No Person Can
Tend’s AI employees run thousands of jobs at once, remember every decision your company has made, negotiate directly with other companies’ AI and turn each call you make into policy. Your people set the objectives and make the judgment calls.
LegalMeet Ava, the AI employee for in-house legal teamsGood morning, Dana.
Legal received 2,418 requests while you were away.
- 2,367complete
- 38waiting on other parties
- 13need your judgment
Ava and 552 temporary specialists worked overnight
- Globex · uncapped GDPR indemnityEU acquisition · 4 precedents
- Initech · data residency in BrazilNew jurisdiction · no precedent
- Northwind · unlimited audit rightsStrategic account · 2 precedents
- and 10 more
Uncapped GDPR indemnity
Globex’s counsel proposed an uncapped GDPR indemnity, which is outside policy. I found four precedents, asked our privacy agent to model the exposure and prepared three positions. I recommend B.
Try it: choose a position, then decide whether it becomes precedent.
Works Inside the Software Your Teams Already Run
- Outlook
- Gmail
- Teams
And any other app your teams open in a browser, with or without an integration.
Capabilities
Abilities No Human Employee Has
Speed is the smallest part of it. An AI employee remembers everything, works in a thousand places at once and tests a decision before making it. Those abilities compound, which is why the gains are exponential.
Perfect recall
Every matter, decision and exception, remembered exactly.
Massive parallelism
Thousands of jobs at once, where a person manages a handful.
Around the clock
Work keeps moving while your team sleeps.
Instant replication
A proven employee can join a new team in minutes.
Continuous learning
Each decision a person makes improves the next thousand.
Machine-speed coordination
Agents settle in seconds what takes people weeks of email.
Specialists on demand
Temporary experts appear for a job and retire when it’s done.
Cross-company networking
Your AI employees negotiate directly with other companies’.
Simulation before action
A change is tested against thousands of scenarios before it ships.
Programmability
When the right tool doesn’t exist, it builds one.
At work
What That Looks Like on a Tuesday
One employee becomes a workforce overnight. Companies close deals with each other in minutes. The software a team is missing gets built by the employee who needed it.
Massive parallelism
One Employee, an Elastic Workforce
A lawyer works a handful of matters at a time. When 2,418 requests arrive overnight, Ava starts the specialists the workload needs and retires them when the work is done.
AAvaAI legal employee
428contract agents
73diligence agents
51privacy agents
2,418tasks running
13need judgment
Dana ReyesGeneral Counsel
I created three temporary agents for the Globex deal: a privacy specialist, a German-language reviewer and a diligence analyst. They’ll be retired when the acquisition closes.
Machine-speed coordination
Two Companies Close a Deal in Nine Minutes
Acme’s AI employees and yours negotiate directly. Sales, legal, security and finance agents settle what they’re allowed to settle, and send the few real trade-offs to the people who own them.
- Elapsed
- 8:57
- Settled by agents
- 87
- Decided by people
- 3
- 10:02:03Acme procurement AIOpens a deal for 400 seats.
- 10:02:04Sales agentsExchange pricing, volume and commercial context.
- 10:02:07Ava ↔ Acme legal AIBegin negotiating the MSA.
- 10:02:19Security agentsSwap questionnaires and audit evidence.
- 10:02:42Finance agentsAgree payment terms and verify billing details.
- 10:03:18Ava ↔ Acme legal AIResolve 87 terms inside both companies’ rules.
- 10:04:02Both sidesSend 3 decisions to the people who own them.
- 10:09:40Dana and Acme’s GCDecide the 3 terms.
- 10:11:00Both sidesContract signed.
It Improves Its Own Workflow
Ava tracks its own turnaround. When a new step slows NDAs down, it designs a faster route that keeps the control in place, and asks before switching.
It Tests a Decision 40,000 Times First
Before Theo changes collections policy, it runs the change against 40,000 simulated customers and shows the CFO the expected cash, the churn risk and the edge cases.
It Spends Within a Budget
With a $250,000 budget, Theo negotiates with vendors, buys software and renegotiates renewals without asking Finance about each purchase. Every dollar is on the record.
Finance Closes the Books Overnight
The CFO goes to sleep. Finance agents close the month, reconcile every entity, chase missing documents and leave the judgment calls for the morning.
It Remembers Every Negotiation
Ask a question no person could answer from memory, and get an answer built from every matter the company has run within Ava’s permitted scope.
Objectives, not tasks
Tell Your Company What You Want to Happen
Founder · Monday, 9:00 AMLaunch our business in Japan.
Tend forms a temporary expansion team of legal, finance, recruiting, sales, procurement and operations agents that work together around the clock.
The founder didn’t operate seven apps or brief seven people. They stated an objective.
Japan expansion72 hours later
- Entity structureProposed
- BankingReady for approval
- Employment agreementsDrafted
- Payroll providerSelected
- Target accounts30 researched
- Candidate interviews12 scheduled
- Local pricingProposed
7 decisions need a person
Review decisionsEarned autonomy
AI Employees Earn Responsibility
Ava starts the way a new hire does, with narrow work and close review. As the work holds up, its manager promotes it by widening what it may do.
Day one
Junior Legal Operations
Autonomy12%
- Handles NDAs
- Every external email needs approval
Six months
Legal Operations
Autonomy68%
- 18,481 matters completed
- 99.3% accepted without changes
- Handles alone: NDAs, standard MSAs, vendor agreements, standard DPAs
Three years
Director, Commercial Legal Operations
- Manages 12 permanent AI employees
- Runs up to 418 temporary agents at once
- Owns commercial contracting, vendor legal and legal intake
Each promotion is a permissions change that Ava’s manager approves, and can reverse.
Ava for in-house legal
Day One Starts With One Contract
Every AI employee starts supervised. On its first day, Ava takes a contract request from intake to sent redline across your legal apps and brings counsel the two calls that need a lawyer.
Matters / ACME-2026-114
Acme MSA Negotiation
Needs your judgment · 2 decisions
- 109 accepted within your playbook
- 15 countered
- 1 cleared by Security
- 2 need you
Work logToday
- 9:02OutlookPicked up Jordan’s request: Acme wants the MSA signed by Friday.
- 9:03iManageOpened the matter and three prior Acme agreements.
- 9:05LegoraReviewed Acme’s 127 changes against your playbook and past Acme deals.
- 9:11IroncladPulled the approved fallback clauses.
- 9:12SalesforceChecked the account. Strategic, $2.4M a year, renewing in Q1.
- 9:14WordAccepted 109 changes your playbook allows and countered 15 with fallback language.
- 9:15SlackAsked Priya in Security to review Acme’s data processing terms.
- 11:40SlackPriya approved the data processing terms. Redline updated.
- 11:42OutlookDrafted the cover note to Acme’s counsel.
Needs your judgment2
Limitation of liability
Acme wants the cap raised from 1× to 2× annual fees. Your playbook allows up to 1.5× for strategic accounts.
Termination for convenience
Acme added a 30-day right to terminate. Your playbook doesn’t offer one. Two strategic deals last year got 90 days.
Choose an answer for each term. Ava has drafted everything else.
Try it: Ava working a contract for an in-house legal team. You’re Dana, Ava’s manager.
What Every AI Employee Gets on Day One
- Its own identityEmail, Slack and a directory account. It never borrows a login.
- Its own computerA persistent browser and desktop, for apps with or without an API.
- Access to your appsScoped like any employee’s, and revocable at any time.
- A job descriptionThe outcomes it owns, written in plain language.
- A managerA named person who approves, escalates and promotes.
- Rules it can’t changeWritten by your team and enforced outside the model.
- A memoryYour playbooks, precedent and every decision since it started.
- A full recordEvery action with its evidence, rule, approver and result.
Authority Grows. Control Stays With You.
Every promotion, budget and permission is granted by a person and can be taken back. The rules sit outside the model, and every action is written down.
Example: Ava’s rules in its first month
On its own
Routine work inside your playbook.
- Read matter files and email
- Run reviews in Legora
- Accept changes your playbook allows
- Ask colleagues for input
- Keep matter records current
After asking you
Anything that leaves the building or needs judgment.
- Send anything outside the company
- Agree to terms outside the playbook
- Change a customer record
Never
Decisions that stay with people.
- Sign agreements
- Edit your playbook
- Move money outside its budget
Every Action, on the Record
Ask why an AI employee did something and you get an exact answer: what it did, why, what it relied on, which rule allowed it and who approved it.
Sent the redline to Acme’s counselThursday, 11:58 AM
- Why
- Acme asked for a signature by Friday
- Based on
- Legora review, your playbook, three prior Acme agreements, Security’s sign-off
- Rule
- Anything sent outside the company needs approval
- Approved by
- Dana Reyes, General Counsel
- Result
- Delivered to Acme’s two counsel
Our thesis
Software still assumes a human operator.
Companies have spent decades buying software for employees to operate. Now AI can operate that software itself. Tend builds the employees, and gives them abilities people can’t have.
- 1990s
- 2010s
- 2020s
- Now
Memory Compounds. That’s the Moat.
Every month on the job, an AI employee knows more about how your company works. A replacement starts from zero.
- Year 1Knows your playbook and how your team applies it.
- Year 3Has seen 60,000 of your negotiations, and why each one went the way it did.
- Year 7Remembers every negotiation the company has run within its scope, including what a retiring colleague was allowed to leave behind.
Software
Priced per seat
Makes one person faster. Judged against other software, paid from the tools budget.
AI employees
Priced on the work
Adds capacity. Judged against the cost of a hire, paid from the headcount budget.
For example: an AI legal employee handling up to 800 requests a month for $8,000 a month.
One Model. Different Jobs.
An AI employee gets a job description, tools, permissions, operating rules and a manager. Change those, and it does a different job.
- AvaLegal operationsLegora · Ironclad · Outlook · iManage
- TheoFinance operationsNetSuite · Ramp · Stripe · Salesforce
- SamRecruiting operationsGreenhouse · LinkedIn · Gmail · Calendar
- MayaProcurement operationsCoupa · DocuSign · SAP · Slack
Get in touch
Hire Your First AI Employee
We’re bringing Ava to in-house legal teams first, and raising our first round to bring AI employees to every operational team.
For legal teams
Bring Ava onto your team, with hands-on onboarding to your playbook and apps and a direct line to the people behind it.
For investors
The deck, a live walkthrough and our plan for every operational team.
Reply to the message that brought you here and we’ll find a time to talk.