Method
How we work.
Evidence before AI. People before automation. Measurement before expansion. Here is what that means in practice.
01 / The method
Five stages, each ending in a decision.
- 01
Understand:
we learn how the work runs today from the people who do it, with real examples.Decision:which workflow matters most.
- 02
Map:
we map each step, take a dated baseline, assign a treatment to every step and test AI on real cases.Decision:build, build with conditions, or not yet.
- 03
Design:
we specify the redesigned workflow, the approvals, the data flow and the measures.Decision:you approve the design and the scope.
- 04
Build and test:
we configure and build in your environment and test against real cases with your team.Decision:ready to launch, or not yet.
- 05
Launch and measure:
we launch with approvals in place, train the people involved and measure against the baseline.Decision:refine, expand or stop.
02 / Four treatments
How we choose between software, AI and people.
Remove
when a step exists only because of a missing question, a duplicate list or a handoff nobody needs. Removing it is the cheapest, most reliable fix.
Software
when the rule is clear and the inputs are predictable: routing, syncing, scheduling, reminders, reports. Ordinary software is faster, cheaper and easier to audit for this work.
AI assist
when inputs vary and the output can be checked quickly: sorting messages, reading documents, drafting replies, summarizing, answering from approved sources. AI prepares; a person confirms.
Person
when a decision carries professional, financial or relationship responsibility, or the cost of an error is high. These steps stay with people, supported by better information.
03 / What stays with people
What always stays with a person.
- 01Commitments to customers, such as prices, quotes, refunds and exceptions.
- 02Advice that requires a license, and anything that binds coverage or a contract.
- 03Payments and financial approvals.
- 04Decisions about people: hiring, performance and access.
- 05Anything published or sent in your name, unless it uses a template a named person has approved.
- 06Every case the system can't handle confidently.
Approval can happen one item at a time, or once for a template that is then used under clear rules. Either way, the approver is named and the approval is recorded.
Tier 1
Runs within rules
Examples: acknowledgments, routing, reminders and status updates from approved templates.
Oversight: a person approves the rule or template once and reviews samples.
Tier 2
Drafted, then approved
Examples: replies outside templates, file notes, document data before it's saved.
Oversight: a named person approves each item.
Tier 3
Decided by a qualified person
Examples: advice, coverage and contract changes, pricing, refunds above a set limit, exceptions.
Oversight: AI may prepare information. It never decides.
04 / Data and security
How we handle data.
Our practices are simple, written down and agreed with you before any build.
Accounts you control:
wherever possible, what we build runs in your accounts and administrative control stays with you.Least access:
we ask only for what a workflow needs, granted through your administrator. At handoff we ask your administrator to remove it and confirm when it's gone.No passwords by email:
credentials are shared only through your organization's approved method.Vendor settings documented:
AI providers' data-use terms depend on the plan and settings you use. We document what's configured for your accounts and point you to the vendor's own terms rather than making promises on a vendor's behalf.Test before real data:
we build and test with samples or test data first, and move to live data only when you approve.A written data map:
for each workflow, what data it touches, where it's stored, which vendors process it and who can see it.Your requirements, reviewed first:
before work begins, we review your vendor-oversight, security and record-keeping requirements with you, say which we can meet and which we can't, and put that in writing.What we don't claim:
FindMeNow doesn't hold security certifications of its own. Where certifications matter, we favor vendors that hold them and record their status in the data map.
Your systems
you own them
The workflow
runs in your accounts
AI provider
your business account, settings documented
Named reviewer
approves what goes out
05 / Measurement
How results are measured.
Each implementation is measured on the same few things before and after launch: how long the work takes from start to finish, how many times it's handled, how often it needs correcting and how much staff time it uses. Measures are dated, taken from your own systems and samples, and reported in plain language, including what didn't improve.
| Measure | How it's measured | Baseline | After launch |
|---|---|---|---|
| Measure: Time from request to completion | How it's measured:From timestamps in your systems | Recorded during Discovery | Re-measured after launch |
| Measure: Touches per request | How it's measured:From a sample of real cases | Recorded during Discovery | Re-measured after launch |
| Measure: Rework and corrections | How it's measured:From review notes and edits | Recorded during Discovery | Re-measured after launch |
| Measure: Staff time | How it's measured:From short time studies with your team | Recorded during Discovery | Re-measured after launch |
- Baseline:
- Recorded during Discovery
- After launch:
- Re-measured after launch
Your numbers, measured, not estimated.
06 / When we say no
When we'll recommend against AI.
- The process isn't stable yet. Fix the process first.
- The output can't be checked quickly enough to be safe.
- The data can't be used with an AI provider under your obligations, and there's no compliant alternative.
- The volume is too low to repay the effort.
- No one can own the workflow after launch.
- A feature you already pay for does the job.
Saying no early is cheaper for everyone. It's also why our yes means something.