AI customer service agents for small businesses
Send repetitive support work over to the machines.
Machine-like builds and runs agents that answer routine requests, update your systems, and hand exceptions to your team.
“My order hasn't arrived. Can you check where it is?”
- 09:41:08InfoOrder and delivery status found
- 09:41:09InfoApproved support policy checked
- 09:41:11DoneTracking link and next step sent
- 09:41:11DoneTicket updated and sources recorded
Your team keeps its attention for the cases that need it.
- 01Start
- $0 to startNo setup or integration fees
- 02Evaluate
- Free 14-day evaluationBegins after build and deployment
- 03Then
- $249 / monthPer workload, starting in Operator Tier
- 04Terms
- Month-to-monthPause or cancel anytime
02Customer proof
A more responsive support operation. Same existing stack.
fitDEGREE brought agents into ticket triage, context retrieval, response preparation, record updates, and escalation.
Read the fitDEGREE story- Tier 1 tickets82%of Tier 1 support tickets reported as automated
- First response68%reported reduction in average first-response time
- Ticket backlog54%reported decrease in ticket backlog
Our partners at Machine-like gave us a more responsive support operation without forcing a massive migration.
How to read these results
These figures describe the published fitDEGREE engagement. “Automated” covers support work and does not establish that every ticket was resolved without human involvement. The case study does not publish ticket counts or absolute before-and-after baselines. Your trial evaluates your own scoped work; these percentages are not promised results.
03Starting points
What is consuming your support time?
Start with work you recognize. These are starting points, not a fixed catalog. We scope the right workflow around your systems and rules.
Morning ticket triage
“There are new tickets in three places. Which ones need us first?”
Your helpdesk or inbox, customer records, ticket categories, and routing rules.
Requests classified, priority set, customer context attached, and exceptions routed to the right person.
Ambiguous cases, complex issues, and approval of replies where agreed.
Correct classification and routing, time spent per ticket, and work that still needs review.
A good first workflow has
- Recurring requests
- Clear policies
- Accessible systems
- Inspectable completion
Mostly unusual or judgment-heavy requests? Start with triage or assisted drafting. We establish what is safe to delegate before the agent acts.
04See the work
An answer is only part of the job.
Follow a request through the source, the action, the customer reply, and the record your team can inspect.
IllustrativeIllustrative workflows with fictional requests. These are not fitDEGREE production records or evidence of measured performance.
A routine request
Path A · Permitted workflow- 10:02:14InfoCustomer request“My order hasn't arrived. Can you check where it is?”
- 10:02:15InfoApproved source & contextThe order record shows it is in transit. The carrier has a tracking link. The approved delivery policy explains the next step.
- 10:02:17DonePermitted actionAttach the order status, share the tracking link, and set a follow-up when the agreed policy calls for one.
- 10:02:18DoneCustomer replyThe customer gets the current status and a clear next step.
- 10:02:18DoneInspectable recordTicket updated with the request, sources consulted, action taken, reply, and follow-up status.
“Your order is still in transit. Here is the carrier's tracking link. If it misses the delivery window in our policy, we'll follow up with the next step.”
- Work happens in your systems.
- The agent retrieves current context and updates the ticket. Your team keeps its existing tools.
- You can see how it got there.
- The source, the action, and the finished output belong in the record. A fluent reply alone is not enough.
An exception reaches a person
Path B · Decision required- 10:06:51InfoCustomer request“I am outside the return window, but I want a refund.”
- 10:06:52InfoApproved source & contextThe order date falls outside the approved return window. Purchase history and the customer's message are attached.
- 10:06:53GatedPolicy boundaryThe agent cannot authorize an exception. The refund action is held.
- 10:06:54InfoHuman handoffA teammate receives the request, order context, relevant policy, and the decision needed: approve an exception or follow the standard policy.
- 10:06:54DoneInspectable recordThe held action, escalation destination, and eventual human decision are recorded.
“This refund falls outside the approved window. The request and order history are attached. A policy exception needs your decision before any refund is issued.”
- The boundary is part of the workflow.
- Missing sources, sensitive changes, and unclear requests reach a person. The agent does not invent a policy to finish a ticket.
- People start with the whole story.
- Context travels with the request so your team can make the decision without repeating the lookup.
FIG. 02 · Two illustrative runs, one routine and one gated. The actions and sources for your workflow are agreed during scoping.
What the agent can own
06 capabilities · scoped per workflow- 01
Grounded answers
Use approved documentation, policies, and current account or order data.
- 02
Ticket triage
Classify, prioritize, and route requests as they arrive.
- 03
Permitted actions
Complete agreed updates and routine actions in connected systems.
- 04
Contextual escalation
Hand judgment calls to a person with the history and policy attached.
- 05
After-hours coverage
Handle scoped requests when your team is away.
- 06
Follow-through
Run the agreed check-ins and follow-ups after the initial response.
05Boundaries
Decide the boundaries before the agent acts.
Your people keep the judgment calls.
You agree what can move automatically, what needs approval, and what reaches a person. We build those rules into the workflow.
Within the agreed scope
Permitted automatically
- Answer covered questions from approved sources
- Classify and route tickets
- Perform explicitly permitted updates
A decision before action
Requires approval
- Actions above agreed limits
- Sensitive account changes
- Exceptions to your standard policies
Context goes with the request
Escalated to a person
- Missing or conflicting sources
- Unclear customer requests
- Situations outside the scoped workflow
Examples, not universal settings. Permitted actions, approval rules, brand voice, and escalation destinations are agreed for your workflow.
06Managed Agent Operations
We build it.
We keep it working.
Managed Agent Operations means Machine-like designs, deploys, and runs your agents as a service. Your team evaluates finished work.
AGENTConnections depend on the workflow and authorized access. We confirm your systems during scoping.
- 01Build & deploy
We translate your process, policies, and completion criteria into a working agent.
- 02Monitor & maintainOn watch
We own production operations and address failures when tools, inputs, or APIs change.
- 03Improve as your business changes
We update the agreed workflow as your policies and support needs evolve.
Your part: provide business context, authorize systems, and make the approvals and judgment calls agreed during scoping. You do not have to build or operate the agent.
07Capacity estimate
What would getting that time back be worth?
Use your own queue to estimate recovered capacity. Your trial replaces these assumptions with observed work.
Estimated capacity recovered each monthEstimate
5 minutes released per eligible request.
- Hours
- 400 requests × 5 min ÷ 60 = 33.3 h
- Value
- 33.3 h × $30/h = $1,000
The example has positive estimated value. Confirm handling time and work quality during your trial.
This values time released, not a promise of lower payroll or measured savings. Include review, corrections, and exceptions in remaining handling time. A negative result is useful: the proposed scope may not justify the cost.
Operator: $249/month per workload. Growth: $199/month per workload from three workloads. Scope and limits are agreed before enrollment. See full pricing →
08Free trial
Agree what success looks like.
Then watch the work.
Your 14-day evaluation begins after we build and deploy the agent. We handle the configuration and operations.
- 01
Create an account
Enter the Terminal and identify work you want to delegate.
- 02
Scope it with us
A required call establishes the workflow, systems, rules, and a correctly finished result.
- 03
Connect & build
You authorize the systems. We build and deploy the agent.
- 04
Evaluate for 14 days
Inspect finished work, handling time, exceptions, and quality in your environment.
- 05
Choose to continue
If the work earns its place, enroll in the agreed service tier.
Your trial scorecard
Starting workflowMorning ticket triage
We agree the targets during scoping. They reflect your work, not another customer's percentages.
- Correctly completed workAgreed during scoping
- Eligible requests handledAgreed during scoping
- Human handling timeCompare before & during trial
- Exceptions & handoffsReview the reasons
- Corrections & reopened ticketsInspect work quality
- Paid scope & priceAgree before enrollment
No credit card or obligation. No setup, integration, or platform fees. If you continue, cancel at any time.
09Questions
Before you hand over the work.
The practical questions about scope, systems, and responsibility.
Q.01What counts as a workload?
A workload is the complete stream of work we agree to own. It can contain multiple workflows, including the supporting updates and reporting. We agree the included actions, volume, and boundaries during scoping. It is not an unlimited promise to handle every support request.
Q.02Will the agent sound like our brand?
We use your approved examples and voice rules, then evaluate the replies with your team during the trial. Requests outside the agreed rules reach a person.
Q.03What happens when it doesn't know?
It does not invent an answer. Missing sources, conflicting policies, and ambiguous requests reach the agreed human destination with the context attached.
Q.04Do we need to change our tools?
We scope agents around your existing helpdesk, inbox, commerce, billing, and customer records. We confirm feasibility and required access for your particular systems before building.
Q.05What access does the agent get?
You authorize the systems required for the agreed workflow. We define permitted actions and approval boundaries during scoping. Review the access and data-handling requirements for your tools with us before connection. Read about security.
Q.06How fast can we start?
Create an account first, then scope the work with us. Build time depends on the systems, policies, and workflow. The 14-day evaluation starts once the agent is built and deployed.
Q.07What does it cost after the trial?
Operator starts at $249/month per workload. Growth is $199/month per workload from three workloads; six or more is a custom quote. We agree your scope and price before enrollment. See full pricing.
Q.08What if the trial doesn't earn its place?
You have no obligation to enroll and do not owe the build cost. Use the completed work and trial scorecard to decide whether continuing makes sense.
Q.09Does this replace our support team?
The first workflow can take repetitive handling, lookups, triage, and follow-up off your team's plate. Your people retain the judgment, approvals, and complex customer conversations agreed during scoping.
Your starting point Morning ticket triage
Give your support team
its time back.
Start with one workflow. See the finished work. Decide whether it belongs in your business.
We build and run the agent. You evaluate the work.