Hire your AI employees
SoClaw is a team of AI employees for your loyalty program. Each one holds a real job — winning back customers, approving orders, protecting points, building campaigns — and handles the complex, repetitive work no human has the time to do one case at a time, all the time.
They work brilliantly, never tire, and are honest about what they achieve — at a fraction of the cost of headcount. You manage them like a team: give goals, set budgets, watch their tasks, and coach them with feedback.
Off by default · you set the limits · pause everything in one click.
Decision: issue a $10 win-back coupon, 30-day expiry.
Held within your spend cap. Logged to the customer's record. Result measured against the untouched group.
Illustrative — figures are an example, not a result.
The work that never gets done, done — one case at a time
Every loyalty program is full of high-judgment, one-at-a-time work that a human team never quite gets to: winning back the customer who's quietly slipping away, checking the receipt that doesn't look right, freezing the points that smell like fraud, shaping the campaign that's been on the list for weeks. It's exactly the work that gets dropped — because it takes patience, attention to each individual case, and the judgment to know when not to act.
SoClaw does that work — as employees you hire.Each AI employee holds a job, works every case with the same care your sharpest specialist would, and does it for all of them, all the time, inside the limits you set. They create outcomes a stretched human team simply can't reach — and they cost a fraction of what the headcount would.
Four AI employees, each with a real job
Every role runs on the same governed footing: off by default, capped, and never confiscating or contacting on a single signal alone — the unclear cases always go to a person. Hire the ones your program needs.
Loyalty Manager
RetentionWorks your high-value customers who are quietly slipping away — one at a time. Picks a single personal move to win each one back, or decides the kindest thing is to leave them alone, and proves the real, incremental revenue against an untouched control group.
- Ranks by where outreach actually changes the outcome
- One bounded offer per customer — points or a coupon
- Proves incremental lift, not vanity redemptions
Order Manager
OperationsClears the review queue no one likes: low-confidence receipts and marketplace orders that would otherwise sit waiting for a person. Approves the clear ones, rejects the bad, and escalates only the genuinely unsure.
- Reads receipts and verified marketplace orders
- Approve · reject · escalate — a decision per case
- Only the unclear cases go to a human
Risk Manager
Risk & fraudGuards your points economy. The moment a rule spots something off — points kept on a refund, a household farming signup bonuses, a burst that looks like wash trading — it freezes those points and decides: void, release, or escalate. It never confiscates on a single signal.
- Freezes suspicious points the instant a rule fires
- Void · release · escalate each hold
- Clears genuine families and bulk buyers, not just punishes
Marketing Manager
Growth · earlyTurns a plain-language goal into a campaign ready to run. Describe what you want — “bring back lapsed VIPs before the holidays” — and it drafts the audience, the offer, the channels and the copy, then hands it back for you to approve and send.
- A goal in plain words → a ready-to-run campaign
- Drafts audience, offer, channels and copy together
- You approve before anything reaches a customer
More roles are on the way. Every new AI employee joins the same console and the same guardrails — nothing runs until you hire it and switch it on.
One case at a time, end to end
Whatever the job, an AI employee doesn't fire blasts at a list. It picks up one case, thinks it through, makes a single fitting decision, records it — then moves to the next. The same care, every time.
Picks up one case
Each employee works from a ranked task list and takes a single case at a time — one customer, one receipt, one hold — starting where its attention changes the outcome the most, not just where the biggest number is.
Reads the whole picture
Before it acts, it gathers everything relevant to that one case: history, consent, how often the customer's already been contacted, the risk signals — exactly the context a careful human would pull together first.
Makes one move — or holds back
It chooses a single, fitting action sized to the case — a bounded offer, an approval, a hold. Often the right answer is to do nothing at all, and every employee is built to choose that too.
Writes down what it did and why
Every decision lands on the record in plain language — the move it made, the reasoning, and what it's watching for next. Your team always sees what happened, and can audit any call.
One case, one move. An AI employee makes at most a single fitting gesture per case — never a flurry of coupons or a string of messages. The goal is the outcome, not the activity.
Manage them like a team
Hiring an AI employee shouldn't mean losing sight of it. SoClaw gives you a manager's console for the whole roster — the same four levers you'd use to run any team.
Tasks
任务Every employee works a ranked worklist, never a black box. Each task is one case with the full reasoning shown and a decision logged. Watch the queue, see exactly why each item ranks where it does, and open any task to audit the arithmetic behind it.
Goals
目标Tell an employee what you want, not how. Hand it an objective — “win back lapsing VIPs”, “keep points-fraud losses down” — and it diagnoses the situation, proposes a concrete plan of moves, and works toward it. You review and approve the plan before it runs.
Budget
预算Each employee works inside a wallet you set — caps per action, per day, per month, plus which customers it may and may not touch. The limits are enforced on Flash's side, not trusted to the employee. When the budget is spent, it stops.
Feedback
反馈You coach them; they never rewrite their own playbook. An employee proposes what it's learning — with the evidence — and a person approves or rejects before anything changes. Course-correct any time, and every action stays on the record to review.
Goals in, budgets set, tasks visible, feedback in the loop — you hold the wheel, your AI employees do the driving you allow.
Autonomy you can actually govern
Every AI employee runs inside boundaries you define, and you can stop any of them cold at any moment. These limits aren't suggestions they choose to follow — they're enforced on Flash's side, so even if an employee wanted to overspend or over-contact, it couldn't.
Off by default
Every AI employee does nothing until you hire it and switch it on. No surprise activity, ever.
You set the limits
You cap what each one can spend — per action, per day, per month — and choose exactly which customers it may work and which to leave alone.
Instant pause
One click stops every action immediately. An automatic safety brake can press the same button for you if complaints start rising.
You approve every change
Employees propose what they're learning — with the evidence — but never change their own playbook. A person reviews and approves before anything new goes live.
Your agent, our guardrails.You run each employee in your own environment with a narrow, capped, time-limited key — only its job, only within your limits. Every spend, contact and budget is checked on Flash's side, not trusted to the agent. Building the integration? The technical contract lives in the Developer Center guides.
They prove the outcomes they actually create
Most tools count activity and call it a win — redemptions, sends, approvals. Those numbers can look wonderful while adding nothing. Your AI employees refuse to play that game: for work that spends or persuades, results are measured against customers deliberately left untouched.
An untouched control group
For work that spends or persuades, a slice of eligible customers is deliberately left alone. Whatever they do on their own is the honest yardstick the AI employee has to beat.
Real, incremental lift — not vanity wins
Success isn't how many coupons got redeemed; that number can look wonderful while adding nothing. What's reported is the extra revenue that only happened because the employee acted, measured against the untouched group.
It flags moves that backfire
Some customers are best left alone — contacting them makes them more likely to leave. Because of the control group, the AI employee can spot those cases and recommend not touching them, instead of quietly wasting spend.
No invented results
It never dresses up a guess as a result or claims a win it can't prove. When a program is too young for reliable evidence, it says so and falls back to honest estimates rather than faking precision.
The number you see is the extra outcome that only happened because the employee acted — the honest one, not the flattering one.
Questions, answered
What is SoClaw?
SoClaw lets you hire AI employees for your loyalty program — autonomous agents that each hold a real job: winning back slipping customers, approving receipts and orders, protecting points from fraud, or building campaigns from a goal. Each works one case at a time inside limits you set and logs every decision. You manage them like a team: give them goals, set their budgets, watch their task queue, and approve or reject what they learn. Think brilliant, tireless specialists that never over-contact, never overspend, and are honest about what actually worked — at a fraction of the cost of headcount.
Which AI employees can I hire today?
Four roles: a Loyalty Manager that wins back at-risk customers, an Order Manager that clears the receipt and order review queue, a Risk Manager that protects points from fraud, and a Marketing Manager that turns a plain-language goal into a ready campaign. The first three are the furthest along; the Marketing Manager is newer. All of them are off by default and rolling out to early-access teams — book a demo and we'll walk through which fit your program.
Do they message customers on their own?
Their autonomous moves are bounded gestures — bonus points, a coupon, an approval — issued within your spend caps and only to customers who've opted in and aren't being over-contacted. When a situation calls for an actual email or text, the employee prepares a draft for a person to review and send; it does not fire free-form messages at your customers by itself.
How do I keep them on budget and in line?
You give each employee a wallet — caps per action, per day, per month — and choose which customers it may and may not touch. Those limits are enforced on Flash's side, not trusted to the agent, so even if it wanted to overspend or over-contact, it couldn't. Everything is off until you switch it on, one click pauses all activity instantly, and no employee ever changes its own approach without a person approving it first.
How do you know their work is actually paying off?
For the roles that spend or persuade, an untouched control group is kept and the extra revenue that only happened because the employee acted is measured against what those left-alone customers did naturally. That incremental lift — not redemption counts — is how success is judged. Every other decision an employee makes is logged in plain language with the reasoning shown, so you can audit any call rather than take it on trust.
Will they annoy my customers?
That's the thing they're most careful about. Before every action an employee checks consent, how recently and often the customer has been contacted, and whether reaching out is likely to provoke a complaint — and it holds back when the answer is yes. If complaints rise across your audience, the whole system pauses itself automatically. Doing less, and doing it right, is the explicit goal.
Why is this so much cheaper than hiring people?
One AI employee does work that would take a team of specialists working case by case — and it works continuously, without the ramp-up, turnover, or overhead of headcount. It's priced as software you switch on, not salaries you carry. Because it's honest about what it does and doesn't achieve, you're paying for real, measured outcomes rather than activity. SoClaw is in early access, so we'll walk through the fit and the terms on a demo.
Is SoClaw available today?
SoClaw is new and rolling out behind a feature flag to early-access teams — it is not a generally available product yet, and we don't publish results we can't honestly back. If you'd like to hire one of these AI employees for your program, book a demo and we'll walk through whether it's a fit.
Only the extra outcome that happened because an AI employee acted, measured against an untouched control group.
Related reading
Keep exploring the pages most related to this one.
Intelligent Decision
A next-best-action: who, which offer, which channel, when.
Read more CapabilityConsumer World Model
The core technology under every decision: a Behavior Model that learns how your members actually behave — trained per brand on their own event stream, honesty-gated against a transparent baseline — plus a Digital Twin that simulates what they'd do next, rehearsing campaigns (and the option of doing nothing) before a message goes out. Together, one complete, living model of your member base.
Read more CapabilityDashboard
Real-time scan-to-redemption data as KPI trends, funnel diagnostics, and a rule + LLM next-best-action panel — plus retention curves by signup cohort and a 14-day revenue outlook that's honest about being a projection, not a commitment.
Read more CapabilitySoTag for Slack
The Slack agent that hands governed, audited numbers to Claude in-thread: daily anomaly scan + human-approved coupon-send. Early access.
Read more CapabilityAI Governance & Approvals
Every AI action runs on a per-action authorization gear — automatic, human review, or forbidden — with an approvals console, a complete decision log, one-click undo, and automatic demotion to human review when an action starts misbehaving.
Read more CapabilityBusiness Ontology
The concept behind the governed AI stack: a world model of your business — objects, properties and relationships plus governed actions — as a metadata layer mapped over the data you already have, so AI knows what a customer is, what has happened, and what it may do about it.
Read more