Driving AI ROI for Atlanta's Mid-Market

The Comprehensive Workshop is one full day of hands-on Claude training for your leadership team and the few people who will drive adoption, at your office and built on your company's own work. It covers model selection and usage, Projects and custom instructions, connectors, and where agents and Claude Code fit, then maps your departments against all of it.

• Two or three of your team's real workflows rebuilt as shared projects, working before the day ends
• You keep the model-selection cheat sheet, the prioritized department roadmap (written up within two business days) and the ground rules for rolling Claude out to everyone else
• A one-hour planning call before the day to pick the real material it runs on, and full recordings after it
• On site at your office, across metro Atlanta and the Southeast

The Working Sessions are hands-on Claude training for the people who run the work every day: three 3-hour sessions about a week apart, for 5 to 25 people in the room plus remote participants. The use cases come from the team itself, and the last session turns the strongest of them into working workflows.

• 01 From AI Chat to Durable Work: what the tools are, where your data goes, and two or three routines each person can run at their desk the next morning
• 02 Your Processes, Augmented: each function maps a workflow it runs every week, builds a first automation against it, and adds what it finds to a prioritized use-case shortlist
• 03 Build Day: working groups stand up three to five use cases from the shortlist in your own accounts, each tested and written up on a runbook card with a named owner
• Hybrid: in the room at your office, with other sites joining live by video. A one-hour planning call and full recordings are included
• A single 3-hour session can also run on its own: a team's first session, one department or one focused topic

Claude Code training for developers and platform teams already using AI tools who need the work to hold up in review. The core is a two-day workshop in which your teams take a real epic from tickets to merged code. A shorter half-day or full-day session covers agents and the controls that make them safe to run.

• Day one, the inner loop: each person takes a guided ticket from research to code review, then builds their own rules files, a skill, a tested subagent and a hook
• Day two, the outer loop: each team slices a real epic into tickets, runs them in parallel, merges working code, and folds one retro fix back into the setup
• Teams of 4 to 8, up to four teams; needs repo access and agent-tool licenses
• Your team keeps the rules files, skills, subagents and hooks it built
• In the shorter session: the controls that live outside the model (access, approval, budget and policy), with observability over all four

A measured answer to where the same AI work can cost less without the output slipping. SZD Labs measures what one unit of a workflow's output costs today, changes one thing at a time, and measures again. A cheaper run counts only if the output still passes.

• A cost per unit of work for each workflow we map, at list price and at your contract price, with a before-and-after figure for every change. Work done in a chat window is allocated, not measured, and the report says so
• The levers: trimming what the model is given to read, pointing it only at the files it needs, tested code for the steps that never vary, a smaller model wherever the answers still pass, and no model at all where plain code will do
• Every change graded against an answer key built from your own signed-off past cases (a workflow needs at least ten)
• Your plan's settings, from the default model to spend limits, checked against the vendor's recommended baseline
• Changes that did nothing are reported too. No saving is promised before it is measured
• How it starts: your exports are read first, then one day on site, then a written report that scopes the work

A use-case-by-use-case answer to whether your AI spend is worth what it costs. SZD Labs builds a ledger with one line per use case. A full line carries what it costs, the unit of work it produces, the before and after, and where the saved time went. The proof is in how each number was obtained, not in a promised return.

• Cost from vendor bills and usage logs, with chat-window cost allocated and labelled as allocated. Every number is graded by where it came from: a bill, a count in your own systems, or what the people doing the work report
• A break-even for each use case with a countable unit: the minutes one unit must save to cover its own cost. A use case with no countable unit is listed, not valued
• Spend cut, seats and tools that can be retired, outside spend replaced and hours returned, kept as four separate figures, with hours counted as capacity, not cash
• Use cases that cost more than they return, reported as plainly as the ones that pay
• How it starts: the same discovery step as AI Token Spend Reduction, on the same cost baseline, so the two can run as one piece of work