AI Automation Guide
How to approach AI automation in your business
Brian Kelly
Founder, Automated Edge
The best starting point for automation is a recurring piece of work your team can explain. A quote assembled from several systems. A report copied into a spreadsheet. Information entered twice. Begin with the work and its cost before selecting AI tools.
Map a complete workflow
Record the trigger, inputs, systems, decisions, handoffs, and final output. Include corrections and exceptions. A five-minute task performed twice a month may not justify a build; a recurring bottleneck across a team may deserve closer measurement.
Ask the people doing the work what slows them down. Agree what measurement is appropriate and tell staff what will be collected. Edgerton, our assessment software, measures the agreed computers for ten business days with that consent in place.
Get a report and a decision
The paid assessment produces findings, ranked opportunities, and a proposal. It describes what we would build, which tools we would connect, the cost, timeline, and estimated payback. You keep the report and decide whether to go ahead. If no clear ROI path is found, the assessment fee is refunded.
Check the assumptions
Suppose a reviewed workflow could recover 5 hours a week at a loaded hourly cost of $40 for 48 working weeks. That is $9,600 in annual capacity value, not automatically $9,600 less payroll. This is an illustrative calculation, not a client result.
For financial return, separate actual expense reductions from incremental contribution after delivery costs. Include assessment, build, software, usage, support, and human review costs over the same period. Allow for rollout time and avoid counting the same hours as both savings and new sales.
Build the agreed work and support it
If you proceed, EdgeKeeper connects the tools you approve on a dedicated server we provide and manage. We build to the proposal, test the outputs and approval steps, and agree acceptance before rollout. Start with a manageable scope and use actual results to decide what comes next.
Once the connections are in place, your team works in Switchboard. Describe a recurring question, request a report or dashboard, or create a read-only AI agent for the job. There is no limit on how many agents you make. Apps and automations that make changes are built with you, with an agreed scope and business case. How the orchestration layer works explains how those agents are limited to the right records and tools.
Support continues after delivery. We maintain the server and connections and help with what we build under your agreement. Pricing follows those stages; no sample example is a fixed package or promise of savings.
Related Articles
How managed AI engagements are priced
Understand the assessment fee, scoped build price, and ongoing service. Compare costs with measured opportunities, not a generic rate card.
How to write an AI use policy for your business
A guide to the decisions behind your AI policy: approved tools, information handling, review, access, and responsible owners.
AI audit for small business: what an assessment should tell you
The evidence, consent, report, and proposal behind a decision to automate. What Edgerton measures and what happens next.
How to detect shadow AI in your business
Four places to look for unapproved AI use, what to record about each finding, and why blocking tools before measuring them makes the problem harder to see.
Is ChatGPT safe for business? What to check
Data controls by plan, the training setting to check, why staff personal accounts are the real gap, and how a written policy and a shadow AI check keep client data where it belongs.
Private AI for business: what private should mean
Four tests for private AI: scoped access, logging, people approving changes, and data you own and can export. What private cannot mean, what it costs, and questions to ask.
Start with the work.
A free 30-minute conversation about your team, tools and repetitive work. We explain whether a paid Edgerton assessment is a useful next step.