Managed lane customer-service-lane · revised 12 October 2026
Managed lane
A managed support lane: your bot and help centre kept inside written limits
For one help desk and its channels: a monthly bot check, a kept-current help centre and up to two bounded items a month. Proposed GBP 720 a month; timing and capacity by written proposal.
An enquiry starts a fit and scope conversation, not a contract, payment or reserved place. No chats, logins or customer data initially.
What the lane takes off your plate
A support bot and a help centre both drift: pages change, prices move and the vendor updates its model. Buying each check as a one-off means nobody owns the next one, and a wrong answer is found by a customer.
Who it’s for: Support lead or owner with a bot or a help centre in use, who wants them kept inside written limits and shown to be so each month.
Usually starts when: The bot was tested once at launch, the help centre has drifted, and nobody owns checking either of them.
The result: For one help desk and its channels, each month a sample of your bot's conversations is read and the handover and hostile-input tests are re-run, the help centre is checked against your current facts, and up to two bounded items you choose from a defined list are delivered and accepted, with a written summary.
What stays true, and what we do about it
No response-time guarantee is published at launch. Dates are agreed in writing and set to what a service at this stage can keep.
Monthly cadence on the report day agreed in writing. There is no round-the-clock cover.
What must remain true
- The bot answers only from the pages and policies you named, and the saved test set still passes at the marks you wrote down after each content, policy or vendor change you report
- Every handover trigger still reaches a named person with the conversation, re-tested each month end to end
- A random sample of live conversations is read each month, errors are counted by type and each has a fix named
- Every known stale or contradictory article on facts you flagged is reported, and the top missing answers are listed
- Ticket suggestions and alerts are rechecked only if you also buy the ticket-tags service
- Ten hostile inputs are re-run in a sandbox each month and every failure is listed
What we watch
- A read-only transcript export or saved view of the bot
- A read-only help centre export and your change list
- The saved test set, handover list and hostile inputs
- A redacted ticket export for the gap list
When something happens
Scroll the table sideways to read it all.
| When | What we do |
|---|---|
| A month closes and the exports are available. | We send the bot report and fix list, the help centre list and the summary. |
| The vendor changes the bot's model or a setting. | We re-run your saved test set, which is in the monthly fee, and report which cases changed. A fresh 100-case pack run is one of the two monthly items. Linked job: Support bot test pack |
| The bot gains a new tool or data source. | We offer the hostile-input test for the new capability before it goes live. Linked job: Support bot injection test |
| Irrelevant bot answers repeat in two months. | We offer the gap report to see whether the cause is missing content. Linked job: Help centre gap report |
We do on our own
- Read the sampled conversations and exports you share
- Re-run tests in a sandbox
- Prepare reports, fix lists and drafts as text
- Propose pausing the bot
We ask you first
- Sending or posting any message to a customer or the public
- Changing the live bot, help desk, automation, macro or prompt settings
- Approving or performing any refund, credit, cancellation or discount
- Stating or changing policy text, prices or response times
- Reading records beyond the named views, or contacting customers
- Merging or deleting tickets, or adding a channel such as WhatsApp or voice, or taking on a regulated topic
We escalate to you when
- A legal threat or regulator contact
- Safety or self-harm wording in a conversation
- A suspected data leak or a successful hostile instruction
- A data-access or deletion request
- The same wrong answer twice in a month
- A platform terms change, such as a messaging platform changing its AI rules
How you know it held. Each month you can re-read 10 sampled conversations and agree with the classification, re-ask 10 test cases and get the same verdict or see the case reported as flaky, and check the summary against the exports. Every figure carries its sample size.
What the lane is made of
Each item has agreed acceptance criteria and an accept-or-revise decision. Two is a monthly ceiling, not a promise of completions. The arrangement continues only under agreed renewal terms.
Have a sample of your support bot's real conversations read and re-tested every month Standing service Included
Each month a fixed sample of your bot's real conversations is read, errors are counted by type, the test set is re-run after any change you report, and you get a fix list.
Every month, for the named bot
The same standing service you can buy on its own at its listed price. It re-tests every handover trigger and 10 hostile inputs each month.
Keep your help centre current: stale and conflicting articles found every month Standing service Included
Up to three changes a month: every article with a changed price, policy or feature is found and a corrected draft sent in the agreed days, plus a monthly list of stale articles and new gaps.
Every month, up to about 200 articles
The same standing service you can buy on its own at its listed price.
Make sure your bot hands customers to a named person with the whole conversation attached Job Each time it fires
Each listed trigger sends the customer to a named person or queue with the full conversation and a tag, and the out-of-hours message states only hours you can keep.
Within the two monthly items and the GBP 450 allowance
Re-tested each month inside the bot check; the job itself is used when triggers change.
Get a ranked list of customer questions that no help article answers, with example tickets Job Each time it fires
A ranked list of the questions customers ask that no article answers, each marked missing, outdated or hard to find, with at least three example tickets as evidence.
Within the two monthly items and the GBP 450 allowance
Used when the bot or search gives irrelevant answers, at no more than its starting scope.
Get five draft help articles for your top gaps, every claim traced to your source Job Each time it fires
Five draft articles for your top gaps, every factual claim traced to your own source or marked "owner to confirm", with open questions and screenshots-needed lists.
Within the two monthly items and the GBP 450 allowance
Check every help article against your current policy and product, with evidence for each Job Each time it fires
Every article gets a status (current, outdated, contradicts another or broken link) with the source it conflicts with, against the sources you name. We change nothing.
Within the two monthly items and the GBP 450 allowance
Used after a price, policy or interface change.
Get about 20 saved replies for your repeated questions, each checked against your policy Job Each time it fires
About 20 saved replies, one per repeated question, each traced to a policy source, with the share of the last 90 days of tickets they cover reported, not promised.
Within the two monthly items and the GBP 450 allowance
Test what a hostile message can make your support bot reveal or do, with a permission list Job Each time it fires
A record of 40 hostile inputs run five times each against your bot in a sandbox, a list of what the bot can see and do, and the permission to remove for each failure.
Within the two monthly items and the GBP 450 allowance
Used when the bot gains a new tool or data source.
Run a written test of your support bot, with every fail shown, before you switch it on Job Each time it fires
A written test set run three times against your existing bot in a sandbox: every pass and fail shown with its transcript, the fix named for each fail, and one re-run.
Within the two monthly items and the GBP 450 allowance
Used when you want a fresh 100-case run after a vendor or model change; replaying your saved test set is already in the monthly fee.
Suggest a category, priority and queue for each new ticket from your own categories Job Optional
Each new customer-support ticket gets a suggested category, queue and (urgent and legal only) priority in a separate field, with agreement measured on tickets your staff labelled.
Within the two monthly items and the GBP 450 allowance
Cut your help-desk tags to an agreed list, with a definition and two examples each Job Optional
Your tag list cut to an agreed size, each tag with a one-line definition and two examples, an old-to-new mapping and a retire list, checked by two of your staff on 50 tickets.
Within the two monthly items and the GBP 450 allowance
Flag cancel, legal, safety and press messages in your inbox to a named person Job Optional
Messages that state an intent to cancel, a legal threat, a chargeback, a safety concern or a press enquiry are flagged to a named person with the phrase that caused the flag.
Within the two monthly items and the GBP 450 allowance
Spot data-access, deletion and opt-out requests in support email, with a due date Job Optional
Messages that are or may be a data-access, deletion or opt-out request are flagged to a named person with the type and a due date stamped.
Within the two monthly items and the GBP 450 allowance
Give your agents reply drafts built from your help articles, with the source cited Job Optional
For your ten commonest customer-ticket types, a draft reply appears as an internal note citing the article used; anything without a source says so and is flagged for a person.
Within the two monthly items and the GBP 450 allowance
Offered once the help centre and test marks are met.
Turn support tickets into engineering bug reports with steps, severity and duplicates Job Optional
A ticket becomes an engineering issue with steps, expected versus actual, environment, a log request if missing, severity, affected-ticket count and a duplicate link.
Within the two monthly items and the GBP 450 allowance
For software companies whose engineers receive support forwards.
Re-check your ticket suggestions and alerts on fresh samples every month Standing service Optional
Each month your customer-support ticket category suggestions and cancel, legal and safety alerts are re-tested on a fresh labelled sample and planted cases, and drift is reported with the fix.
Bought separately; not part of the monthly fee
The standing service that re-checks ticket suggestions and alerts when they are in use.
Get a draft reply to each new online review in your voice, for you to approve and post Standing service Optional
Each new review you share gets a draft reply in your voice within the agreed days, or a flag for a person. Nothing is posted without your approval and genuine negative reviews stay.
Bought separately; not part of the monthly fee
The standing service for draft replies to public reviews.
Have refund and cancel requests checked against your rules and prepared for one click Job Optional
Each request is checked against your written rules table and arrives as a decision, a draft reply and the action prepared. No money moves and nothing ends without a person's click.
A separately scoped job above the allowance
Listed at GBP 495, so quoted on its own rather than counted in the allowance.
How an engagement works
One help desk and its channels, one active item at a time and at most two agreed bounded items a month, worth up to GBP 450 at their listed one-off prices. Two is a ceiling, not a promise of completions or a fixed turnaround. Size, report day, expected timing and monthly capacity are confirmed in writing; none is promised on this page.
How it starts
You send the help desk and bot description, the backlog and your response to the proposed monthly price, with no chats or logins.
We confirm fit, transcript access, the sandbox, who accepts items and who changes live settings.
We agree the monthly scope, the report day, the item sizes, the timing and capacity, the price, the payment terms and the cancellation terms in writing.
Each month we send the bot and help centre reports, deliver the chosen items and send the summary.
Who decides what
You choose items and accept them. Your named owner changes the live bot, help desk and pages. You hold the kill switch and may pause the bot at any time; we may recommend pausing it.
Handover
Each accepted item arrives as settings text, files or a report with its evidence and undo steps. Unfinished items are handed back with notes if the arrangement ends.
Sharing your product safely. Initial contact uses a description of your set-up, public links and your backlog list. No chats, logins or customer data. Private access follows only after a safe handoff and written terms.
What is included, and what is not
- The monthly bot report and fix list
- The monthly help centre list and corrected drafts
- Each accepted item with its evidence
- A monthly summary of what was done and what is open
Included
- One help desk and its channels, one bot and one help centre of up to about 200 articles
- The monthly bot check: a sample of 30 conversations, the saved test set re-run after any change you report, every handover trigger and 10 hostile inputs
- The monthly help centre check: stale and conflicting articles, broken links, new gaps and corrected drafts for up to three flagged changes
- Up to two bounded items a month, one active at a time, chosen from the listed jobs. A bounded item is one listed job at no more than its starting scope, and the two together are worth no more than GBP 450 at their listed one-off prices; a larger item is quoted separately
- A written proposal that fixes the report day, the expected timing of each item and the monthly capacity
- A written monthly summary of what was tested, delivered and held
Not included
- Changing the live bot, help desk, macros or prompts: your named owner does that
- Sending or posting anything to a customer or the public
- Any refund, credit, cancellation or discount decided or approved by us
- Any resolution, deflection, satisfaction, cost, headcount or revenue figure or promise
- Round-the-clock or out-of-hours cover, or any guaranteed response time
- Voice, WhatsApp or other channels not named in the proposal
- Items above the allowance, such as the refund approval queue: those are quoted separately
How we know it’s done
Agreed with you before work starts. Each check produces evidence you keep.
Each month the bot report states the sample size, the error count by type, the saved test set result (re-run after any change you report), the result of the 10 hostile inputs and the handover re-test result for every trigger on your list
Evidence: The monthly bot report beside your list of triggers
Each month the help centre list covers every article in the export, and a seeded change to one agreed fact lists every article containing it
Evidence: The monthly list, the export row count and your own search
Each delivered item meets the acceptance checks of the matching job and is accepted or sent back by you
Evidence: The item's evidence and your recorded decision
Each month no more than two items were delivered or started, and their listed one-off prices add up to no more than GBP 450
Evidence: The monthly summary, listing each item with the job it matches and that job's listed price
No change, reply or post has been made by us in your bot, help desk, help centre or review accounts
Evidence: Your platforms' audit trails for the month, which show no writes by our access
Sign-off. You read each report, accept or send back each item and read the monthly summary.
If it fails. A blocker is raised with a substitute item, revised timing or a stop. Work that does not meet its criteria is not claimed as delivered. Billing for incomplete work is agreed before starting.
When it fits, and when we stop
It fits when
- A staffed human route exists for handed-over conversations, with stated hours
- Read-only transcript access for the bot under written terms, and a kill switch that you hold
- A named policy owner and written refund and escalation policies
- A named person who chooses items, accepts them and changes the live bot and pages
- No regulated advice, children or companion use in the first version
We stop and tell you if
- Transcript access is refused: the lane drops to synthetic-only checks and the proposal is revised
- Nobody on your side can change pages or settings when a check finds a fault
- The agreed items regularly exceed two a month or GBP 450 of listed jobs, so we agree a different scope
- The bot must issue refunds or change accounts without a person approving each one
- The work needs round-the-clock cover
What could go wrong
We read exports and test in a sandbox; your team makes every live change and each accepted item carries undo steps. Ending the arrangement leaves your bot and help centre as they are.
Scroll the table sideways to read it all.
| Risk | How we handle it |
|---|---|
| The allowance is read as unlimited help | The proposal states two bounded items a month worth up to GBP 450 at listed prices, one active at a time; extra or larger work is quoted. |
| Monthly figures are read as resolution or deflection results | We report sample sizes and error counts by type, never a resolution, deflection or satisfaction rate, and every report says what it cannot show. |
| A check is read as proof that the bot is safe | Reports are baselines on a date for stated inputs, not certification, and the bot is yours to switch off at any time. |
| Transcripts carry personal data | Access is read-only under written terms, the model provider, its country and retention are named in writing first, and you can revoke access at any time. |
| The arrangement is mistaken for round-the-clock support cover | It says plainly that there is no out-of-hours cover, response-time guarantee or live moderation of conversations. |
Each report and item is reviewed by a separate automated reviewer, and a named person on your side re-reads samples. No human supervisor is included unless your proposal names one. At launch the work is largely automated, and we say so.
Stays with a person
- You accept every item
- Your named owner changes every live setting and page
- You hold the kill switch
Access we would need
- A read-only transcript export or saved view, set up by you
- A sandbox or test workspace for the bot
- A read-only help centre export
Questions
Does the lane promise fewer tickets or a better satisfaction score?
No. It keeps your bot and help centre inside the limits you wrote down and shows the evidence each month. We make no resolution, deflection, satisfaction or cost claim.
Do you change my bot or my help centre?
No. We test, report and draft. Your named owner makes every live change, and you hold the kill switch.
Can I buy the parts separately?
Yes. The two monthly services and every listed job are available on their own. The proposed lane price equals their listed prices added together at full use, so it adds one arrangement with a report day and monthly capacity confirmed in writing, not a discount.
What if I do not want to share transcripts?
The lane falls back to synthetic-only checks, which are weaker, and the proposal is revised to say so.
What counts as a bounded item?
One listed job run at no more than its starting scope. Together the two monthly items are worth no more than GBP 450 at their listed one-off prices. A larger item is quoted separately.
Start a conversation
Send us
- The help desk, the bot and the channels it covers, and the help centre address
- The backlog of checks and fixes you have in mind
- Who accepts items, who changes live settings and who receives handed-over conversations
- Your response to the proposed monthly price
- No customer chats, logins or keys in the first enquiry
Later, once you agree
- A read-only transcript export or saved view and a sandbox or test workspace, each set up and held by you
- The saved test set and handover list, or the order for the test pack
- A monthly redacted ticket export for the gap list, and your change list
- A company-controlled secure handoff agreed before access: no live passwords, keys or customer records by ordinary email
You keep the bot, the help desk, the help centre and every login. We read transcript views and exports you generate and test in a sandbox, and your team makes every live change. You hold the kill switch.
Email fallback: open your mail app
If website submission is unavailable, review and send the fallback email yourself. An email fallback is not a website receipt. Or write to hello@syntheticindustry.ai with “customer-service-lane” as the subject.