Snaz Consulting Inc.
Delivery catalog · 2026
AI Workflows For Project-Driven Businesses
I build AI systems that give your team back the hours it loses to repetitive admin. Every workflow ends with a person approving the output before anything reaches a client.
How The Work Is Built
Six standing rules
Generate, Check, Approve
Nothing publishes itself. Every system ends with a human reading the draft and saying go. The draft is free; the approval is the job.
One Source Of Truth
The AI drafts from your approved templates and your real data. It does not invent a fact, a number, or a price.
Broad Read, Narrow Write
Agents read widely across your tools. Write access stays deliberately narrow, and nothing goes out externally without approval.
Automate What Repeats Weekly
Frequency decides what gets built, not how irritating a task feels. A monthly one-hour job is twelve hours a year. That stays manual.
Fail Loud
Every scheduled job reports success or failure to Slack or Telegram. A silent failure is worse than no automation at all.
Handover, Not Dependency
You own every account, prompt, template, and check. A new hire could run it from the documentation. Nothing is held hostage.
Systems Already Running
Built in house, operating now
Everything below runs my own business. I built these for Snaz first, which is why I can show you the whole thing rather than a summary: the prompts, the gates, the failure alerts, and what broke before it worked. Any of them can be walked through live on a call.
The Content Engine
- The Problem
- Publishing consistently while running a business is a scheduling and quality problem, not a writing problem. Output either stops or drifts off-voice.
- What Was Built
- A scheduled research task runs weekly, searches the last seven days, drafts two options, and posts them to a review board. A voice-and-style gate checks every draft against a written rule set. Approved posts get an image, then queue for publishing on a fixed calendar.
- The Result
- Running since early July. Content produced and scheduled through week eight, all of it sounding like me, with nothing published before a human approved it.
Sold as Status And Weekly Reporting and Managed AI Operations
The Gated Document System
- The Problem
- A document task I repeat constantly took two to three hours every time, and most of that was assembly rather than judgment.
- What Was Built
- Drafting against an approved template from a single source of truth, then two audit passes: one for unsupported claims and banned phrasing, one for formatting and number verification. A person approves before anything is used.
- The Result
- About 20 minutes a run instead of two to three hours. Across 25+ runs, roughly 60 hours returned against a few evenings of building.
Sold as Document And Proposal Drafting System
My Own Context Pack
- The Problem
- Every AI task started from zero. Output was competent and completely generic, so editing it back into my own voice cost more than writing from scratch.
- What Was Built
- One document holding what the business does, the services and pricing, the templates, the voice rules with specific banned words and phrasings, the definitions, and the never-list. It is fed in at the start of any task.
- The Result
- Drafts land in-voice on the first pass. The same document now governs every system I run.
Sold as Context Pack
The Approval And Governance Layer
- The Problem
- Automation that can send things is automation that can embarrass you. The risk is not bad drafting, it is a bad draft reaching someone before a person sees it.
- What Was Built
- A written rule set every system obeys: read access broad, write access narrow, no external send without approval, and a never-list of things the AI does not touch. Every scheduled job reports success or failure, so a silent failure cannot hide.
- The Result
- Multiple systems running unattended for months with no unreviewed output ever reaching an audience.
Sold as AI Governance And Shadow-AI Review
Any of these can be walked through live, including the prompts, the gates, and what broke before it worked.
Book a callThe Catalog
Select a service to see the build
AI Workflow Audit I sit in your operation and map where AI genuinely saves hours, where it must never go, and what to build first. You get a written plan you own whether or not you build with me. The build
How It Runs
- A short intake questionnaire: what your team repeats, which tools you run, where the week goes.
- A 60 to 90 minute working session watching real tasks happen, not a slide deck.
- Every repetitive task mapped by frequency and hours, then scored on frequency, hours saved, and risk.
- Each opportunity sized for build effort, so nothing is quoted blind.
- The written plan and a recorded walkthrough, delivered to you.
You Receive
- A written plan in Word or Google Docs
- A recorded walkthrough of the findings
- A ranked build order with effort against payback
- A never-list: where AI does not belong in your business
- Half the fee credits against a six-month retainer if you build
Context Pack Generic AI output is a context problem, not a model problem. One document covering what your business does, your services and pricing, your templates and voice, your definitions, and your never-list. It makes every AI output afterwards sound like your business. The build
How It Runs
- An owner interview: what the business does, for whom, services, pricing, voice, definitions.
- Capture the never-list, covering what AI must not touch or claim.
- Assemble everything into one clean document.
- Test it against three real tasks and tighten wherever output still drifts.
- Hand it over as an asset your whole team reuses.
You Receive
- One living document in Google Docs or Notion
- Your templates and tone captured in writing
- A definitions list, so your words mean your things
- A never-list for the whole team
- Tested against your real work before handover
AI Governance And Shadow-AI Review Right now someone on your team is quietly pasting company information into a free chatbot. This closes that gap with an approved-tools list, a never-list of data that touches no AI, and one written workflow showing the safe way to do your most common task. The build
How It Runs
- Inventory current AI use across the team, including the unsanctioned free-tool usage.
- Risk-rank your data categories: fine, sensitive, never.
- Draft the approved-tools list, naming which AI, which accounts, which paid tiers with real data controls.
- Draft the never-list.
- Write one safe-workflow procedure for your most common task, then walk the team through all three.
You Receive
- An approved-tools list with account and tier guidance
- A never-list of data that touches no AI
- One written safe-workflow procedure
- A team walkthrough of all three
- Built on delivery experience under ISO 27001 and HIPAA
Meeting Recap And Action-Item Loop Same-day follow-through. Rough notes or a recording become decisions, actions, owners, and dates in your format. A person scans and corrects, the recap goes out the same day, and the next agenda starts from whatever is still open. The build
How It Runs
- Pick the capture method: a recording with transcription, or pasted rough notes. Messy is fine.
- Build the extraction into your standard format: decisions, actions, owners, dates.
- Add the human scan step, the one place judgment is required.
- Send the recap the same day.
- Seed the next agenda automatically from the still-open items.
You Receive
- Recaps drafted into email, Slack, or Notion
- Decisions, actions, owners, and dates every time
- A running open-items list that carries forward
- Your next agenda seeded automatically
- Calibrated on three to five real meetings before handover
Document And Proposal Drafting System The twenty-minute document. Work that took an afternoon now takes about twenty minutes: the AI drafts against your approved template from a single source of truth, two audit passes check it, and a person approves. The build
How It Runs
- Build the source-of-truth document so the AI never invents a fact.
- Fix the template every draft must follow.
- Configure the drafting agent to write only against that template and source.
- Build two gate passes: unsupported claims and banned phrasing, then formatting and number verification.
- Wire the approval step, then run three to five real tasks live to calibrate before handover.
You Receive
- A drafting system inside the tools you already pay for
- A source-of-truth document the AI cannot deviate from
- Two automated audit passes on every draft
- Number verification before anything is approved
- Documentation a new hire could follow
Status And Weekly Reporting The status lives where the work already happens. An agent reads your project tools and sheets, drafts the weekly update against your template, and flags what changed. You correct and approve in fifteen minutes instead of half a day. The build
How It Runs
- Connect read-only to the systems where status already lives.
- Define the report template in the executive format: where we are, the risk that matters, the decision needed and by when.
- Schedule the weekly pull.
- Add the change-since-last-week flag, so nothing hides inside "in progress".
- Route the draft to you; on approval it goes to the team. A fail-loud alert fires if a source is unreachable.
You Receive
- A weekly draft assembled from your own systems
- Read-only connections, so nothing can be altered
- A change flag on everything that moved
- Slack or Telegram alerts on completion and failure
- The executive report format, built from delivery practice
Owner Inbox Triage The owner's inbox is where small-business hours quietly disappear. The agent sorts mail into needs-you, needs-anyone, and needs-nothing, drafts replies for the middle lane from your context pack, and hands you a short brief for the top lane. The build
How It Runs
- Define the three lanes and the rules that sort mail into them.
- Connect the inbox with read and draft permission only, never send.
- Build the context pack so drafted replies sound like your business.
- For the needs-you lane, generate a short brief: what they want, the history that matters, a suggested answer.
- You review and send. Nothing leaves without that.
You Receive
- Three-lane sorting on incoming mail
- Drafted replies waiting in a review queue
- A short brief on anything that genuinely needs you
- Read and draft access only, never auto-send
- A ping when the brief is ready
Quote And Proposal Automation In services and construction the fastest credible quote usually wins the work. Past proposals become templates, the AI drafts from the enquiry details and your pricing, and a person checks every figure before it leaves. Days become hours; the pricing stays human. The build
How It Runs
- Gather past proposals and your current pricing structure into one source.
- Build the proposal template from the winning past examples.
- Configure the drafting agent to assemble from enquiry details plus pricing. Assembly only; it never invents a price.
- Build the number-verification gate, so every figure surfaces for human approval.
- You approve the pricing and send. This one touches revenue, so the gate is strict.
You Receive
- Quotes drafted in hours instead of days
- Consistent pricing structure on every proposal
- A verification gate isolating every figure for sign-off
- An alert when a new quote is ready for review
- The AI assembles the quote; you decide the price
Connected Agent Stack A chat window answers questions; a connected agent does work. The agent reads your notes, files, email, and calendar, drafts against your systems, and passes every outbound action through an approval gate. Read access is broad, write access is narrow. The build
How It Runs
- Map the systems the agent needs to read and the few it needs to write to.
- Stand up the connections with broad read access.
- Scope narrow, explicit write permissions, nothing beyond what the workflow requires.
- Build the approval gate, so no outbound action fires without a human.
- Add fail-loud monitoring, then roll out in phases.
You Receive
- An agent working inside your real tools
- Broad read access, narrow and explicit write access
- An approval gate on every outbound action
- Logging and fail-loud monitoring throughout
- A phased rollout, not a single switch-on
Managed AI Operations The umbrella that runs and improves everything above. I operate the live workflows, deliver one new or improved automation every month, hold a monthly operations review, and send a monthly statement showing exactly what your systems did and the time they returned. What is included
Each Month
- Operate and monitor every live workflow with fail-loud alerting.
- Deliver one new or improved automation against the audit roadmap.
- Run an operations review with you.
- Send the time-saved statement and update your dashboard.
- Re-prioritize the roadmap as the business changes.
You Receive
- Monitoring and fail-loud alerting on every job
- One new or improved automation monthly
- A monthly operations review
- A time-saved statement and live dashboard
- Half of any setup fee credited on a six-month term
Fractional AI Delivery Lead For an organization scaling its AI use. Two to four days a month embedded in your operation, owning the roadmap, governance, vendor decisions, and team training. The strategic seat rather than a single workflow. What is included
The Engagement
- Embed two to four days a month inside your operation.
- Own the AI roadmap and the governance framework.
- Lead vendor and tooling decisions.
- Train your internal team so the capability stays in house.
- Re-prioritize quarterly as the business scales.
You Receive
- An owned AI roadmap, reviewed quarterly
- A governance framework your team can apply
- Vendor and tooling decisions made on evidence
- Internal team training
- Oversight of any build work the roadmap calls for
How An Engagement Runs
Four stages
Audit
I sit in the operation and map where AI saves hours, where it must never go, and what to build first. You get a written plan you own either way.
Plan
Every opportunity scored on frequency, hours saved, and risk, then sized so the build effort is known before anything is quoted.
Build
Built inside the software you already pay for wherever possible, then calibrated against three to five of your real tasks before it goes live.
Handover
You own the accounts, prompts, templates, and checks, documented well enough that a new hire could run the whole thing.
The Delivery Record Behind It
Why the method exists
Over twelve years I have delivered projects across construction, e-commerce, insurance, and healthcare technology. The discipline came first. The AI systems are built on it.
150+
Healthcare organizations taken through EHR implementation and migration, supporting 12,000+ patients
PointClickCare
50+
Concurrent projects run at once, with executive steering cadence across business, IT, and security
PointClickCare
50+
Databases migrated to Oracle Cloud with full data integrity and minimal downtime
Medavie Blue Cross
$1.65M
Eight-unit residential development delivered on time and on budget, coordinating 25+ subcontractors
Mc Ahad Ltd
Certified PMP with the Project Management Institute and PSM with Scrum.org. Compliance delivery experience under SOC 2, ISO 27001, and HIPAA-governed environments.
Standing Terms
No surprises
- You Own Everything
- Every account, prompt, template, and check is yours at handover. There is no platform to stay subscribed to and nothing held back.
- Setup Credits Toward A Retainer
- Setup is a real fee, and half of it credits against invoices on a six-month retainer.
- Three-Month Minimum On Retainers
- Workflows need a quarter to prove themselves properly. Anything shorter measures noise.
- Prerequisites Are Quoted Separately
- If a workflow needs something you do not have yet, a CRM, a shared source of truth, a template store, standing it up is a separate line rather than absorbed into the build.
- Always-On Workflows Carry A Small Subscription
- A workflow that runs unattended around the clock needs an orchestrator account in your name, roughly $20 to $25 a month, paid by you directly.
- Priced By Result
- Fees are set against the outcome and the scope, not against hours at a keyboard.
Start With The Audit.
One working session, a written plan you keep, and a ranked build order. If nothing on the list pays for itself, that is a useful answer too.
Snaz Consulting Inc. · Calgary, Alberta · All prices in US dollars and current for 2026