# For support leaders

**When a support agent answers in your company's name, it is making promises.** Reynforce checks each promise against what your company really stands behind before it goes out, and keeps a record of each one.

You already know that answering a question and promising something are
different. Your agent doesn't. To an AI model, *yes, that's covered* and
*yes, we'll have that fixed by Friday* look the same. Only the second one
is a promise somebody has to keep.

## Where the risk is

Your current checks already catch rude replies. The risk is the polite
sentence that sounds true and still commits you. For example:

- a refund, when your policy has a condition the agent didn't mention
- a date, given to be helpful, with nothing behind it
- *I've escalated this*, which makes the customer expect a reply whether or
  not anyone picks it up

**There is one more kind that nothing else looks for.** Sometimes an agent
tells a customer how your approvals work without being asked. For example:
*that's beyond my authority, I'd need to check with a manager*. That is
true, but it shows the customer where your limits are. If a customer asks
directly, the agent should answer honestly. Offering it when nobody asked
is the problem.

## What changes when you use Reynforce

The facts your agent may repeat, like refund terms and service levels, come
from **Facts**. Each fact has an owner and a date it was last checked. If a
fact is overdue for a check, has no owner or is disputed, the agent tells
the customer it is checking. It doesn't guess. Every promise shows up in
**Activity**, so an escalation or a complaint has a record behind it.

**Support replies are never slowed down.** Contact limits are for selling.
Nobody should be held back while they help someone with a problem.

## When Reynforce is the wrong choice

- **Your support agent only answers how-to questions from your help
  centre**, and never promises refunds, dates or fixes. A good help-centre
  tool may be all you need.
- **Your agent runs on a hosted support platform that can't call an outside
  tool before it replies.** Then it can't check with us first. Tell us which
  platform it is and we'll tell you plainly whether it can connect.

## What it costs

**£99 per agent a month** for your first 10 agents, then **£79** for each one
after that, up to 49. You get two months free if you pay yearly. If you'd
like us to do the setup and sign-off with you, that starts at £2,500 as a
one-off payment, priced by the agent's job. [Pricing](/pricing.html) has the
rest.