Field notes on making AI work in real businesses.
Field notes from Nezam AI: essays on shipping AI in real businesses.
- Small business AI mostly breaks between the tools you already pay for
AI consulting for small business is mostly connecting tools you already pay for: where ours failed quietly, and what to ask a consultant first. (2026-10-22) - Ontario businesses with 5 to 19 employees went from first to last on AI use
Small business AI adoption by size, from Statistics Canada: Ontario firms with 5 to 19 employees led on AI use in 2024 and came last in 2026, at 9.2%. (2026-10-22) - Toronto's no-heat complaints arrive with the first night below 7°C
Toronto heating season data: in four autumns, 311 no-heat requests ran five to seventeen times busier the week of the first night below 7°C. (2026-10-17) - An AI receptionist only knows what your office wrote down
AI receptionist setup starts with policy: the 100 calls our live agent had no rule for, and the worksheet we now fill in before go-live. (2026-10-17) - AI intake for Ontario law firms: what the LSO rules let an agent say
AI intake for Ontario law firms, read against the LSO rules and By-Law 7.1: what an agent may not say, and four times our own agent stated the law. (2026-10-07) - Work order completion time: the jobs that wait on an owner
Work order completion time across 2,363 repairs at a Toronto property manager: a median 14.9 days when the owner had to approve, 5.8 when not. (2026-10-07) - Rentvine API: the filters it ignores, and why that matters once AI reads your books
The Rentvine API drops parameters it does not know and still answers 200. What we measured on a live ledger, before Rentvine's MCP reaches every account. (2026-09-30) - Software for plumbers who serve property managers: what we priced, and what we built
Software for plumbers who serve property managers starts with the work order email. We priced five platforms for a GTA plumber, then built a small app. (2026-09-30) - Jacob Coxon had help. He should have said so.
A PR firm was booking Jacob Coxon the day he told Fox News he had no outside help. What the report shows, what it doesn't, and what we got wrong. (2026-09-27) - AI receptionist cost per call: 867 answered calls, itemised from the bill
AI receptionist cost per call, from the platform's own bill: 867 answered calls at 27 cents on average. The model was a fifth of it. (2026-09-26) - The fair housing AI leasing assistant: what it may tell a renter, and what goes to a person
A fair housing AI leasing assistant gives every caller the same facts about the unit. The statute, HUD's 2024 guidance status, and the rules we build to. (2026-09-24) - Massachusetts heating season: which no-heat calls wake the on-call tech
The Massachusetts heating season runs September 15 to May 31 at 68°F by day and 64°F at night. Which no-heat calls wake the on-call tech, hour by hour. (2026-09-24) - TCPA AI lead follow-up: which step needs which consent
TCPA AI lead follow-up, step by step: an AI answering a lead makes no call, but the AI callback and drip texts need consent the lead may not have given. (2026-09-24) - Massachusetts call recording law and the AI receptionist: what the first seconds must say
Massachusetts call recording law turns on secrecy, so an AI receptionist that announces the recording first stays clear of it. The statute, cases, script. (2026-09-24) - LeadSimple API field notes: what it ignores, and when Rentvine work orders really arrive
LeadSimple API field notes from a live account: filters it ignores, what it cannot write, and why Rentvine work orders made after 11 PM wait until 2 AM. (2026-09-23) - AI receptionist mistakes: what our voice agents got wrong in 3,358 graded calls
AI receptionist mistakes, counted: we graded 3,358 calls on two live voice agents. Six emergencies went down as routine and the grader caught none of them. (2026-09-23) - Rent collection automation in Ontario: the PAD rules, and where the run breaks
Rent collection automation in Ontario: why RTA s.108 bars requiring PAD, what Payments Canada Rule H1 demands, and six places a debit run breaks. (2026-09-23) - CASL real estate rules for the AI that texts and calls your leads
CASL real estate rules, read off the statute: which leads an Ontario brokerage's AI may text, email or call, on what consent, and for how long. (2026-09-23) - Real estate brokerage call routing: 41% of callers are other realtors
Real estate brokerage call routing, from 1,190 answered calls: 41% of callers were other realtors, 92.8% of them left a number, and 12.2% were junk. (2026-09-23) - Property management call volume by hour, and who belongs on the phones
Property management call volume on one Toronto line: 2,142 answered calls, a Monday 11am peak, angry first hours and a 5pm cliff. When to staff, and who. (2026-09-23) - Your website says 24/7 and your office closes at five. Check where the calls go before you buy an AI receptionist.
AI receptionist for contractors: the seven checks we run before quoting one, drawn from a Toronto roofing company's website, and where the answers land. (2026-09-22) - The LTB Order Catalogue, read: half of Ontario's tenancy files this year are about unpaid rent
The LTB Order Catalogue, read end to end: 49,269 orders from January to June 2026, half of them arrears, and what 220 of those orders actually decided. (2026-09-21) - After-hours maintenance triage rules: what woke someone in 158 emergencies
After-hours maintenance triage rules from a live line: what got the emergency tag in 2,142 calls, the Ontario heat floor, and the three water questions. (2026-09-21) - We asked three AI assistants for a vendor. Here is what they searched first.
What AI assistants search before they recommend a vendor: ten buyer prompts through ChatGPT, Perplexity and Gemini, every search and citation logged. (2026-09-21) - Ontario's RTA changes land September 21. Most of them are date arithmetic.
Ontario's RTA changes take effect September 21, 2026: a seven day N4, a conditional N12, and a prescribed test for persistent late payment. (2026-09-16) - Claude Fable 5.1 and GPT-6 Astra: what we use each for
Same $10 and $50 list price, different bills. What the cache rate, the 272K cliff and the vendor benchmarks mean once you deploy one of these. (2026-09-15) - What we can promise you the week Jacob Coxon quit
Nobody has asked us about Jacob Coxon yet. They will. Where the frontier warning ends, what an operator controls, and why we want the models open. (2026-09-14) - What 3,332 answered calls say about after hours
Two phone lines at one Toronto operation, every call answered and logged. 19% arrived after hours, and the line the 65% claim is about ran 12.7%. (2026-09-11) - Yardi automation: what your licence reaches
Yardi licenses integration per interface and its partner program has a two year bar. Three lanes decide which Yardi workflows you can build now. (2026-09-11) - Ten AI tools we actually use for marketing in 2026
The ten AI marketing tools open in a working week at a small firm, what each replaced, what they cost in September 2026, and the four we do not pay for. (2026-09-10) - What is AI consulting, actually?
AI consulting defined from inside the work: what the term covers, where the hours go, the four things sold under one name, and when you do not need it. (2026-09-03) - AI consulting cost in Toronto: what the market charges
What AI consulting costs in Toronto across the tiers, why one brief comes back ten times apart, and the Ontario line items a template quote leaves out. (2026-08-31) - AI consulting firms in Toronto: who actually does what
A tiered map of fifteen AI consulting firms serving Toronto, each verified on its own site, plus the five questions that expose a weak one. (2026-08-31) - Follow Up Boss automation: what to wire up first
Six Follow Up Boss workflows worth building, with the trigger, the steps, and the human gate for each, grounded in the real API and webhooks. (2026-08-31) - Clio automation: what a Canadian firm can safely wire up
Five Clio workflows a Canadian law firm can safely automate, with the trigger, the steps, and the human gate for each. None of them produce advice. (2026-08-31) - AI receptionist for plumbers and HVAC: what it actually catches
A trades business loses jobs between the ring and the callback. What an AI receptionist actually catches, and what it should never be allowed to do. (2026-08-25) - AppFolio automation: what connects, and what does not
AppFolio automation, mapped honestly: what Realm-X already covers, what Stack connects, and what needs a custom build around the tiered API. (2026-08-25) - Buildium automation: six workflows worth building
Six Buildium workflows worth automating: the trigger, what the system does, and where the human gate sits. From a team that builds and runs them. (2026-08-25) - An AI answering service for property management: what it costs
What rings a property management line, what an AI agent handles end to end, what it hands to a person, and what it costs next to per-minute services. (2026-08-25) - How AI helps property managers respond to tenant requests faster, and the one clock it cannot move
AI cannot make a repair happen faster. It can compress acknowledgement, triage and dispatch. Peter Lohmann's data shows what moves and what does not. (2026-08-23) - How AI is changing home buying and selling, and where it actually breaks
AI has not replaced the real estate agent. It moved the moment the agent is needed. What the data says, what TRESA requires, and where deals are lost. (2026-08-23) - The real return on AI, and why we will not give you a percentage
Three kinds of return exist, and only one is countable before you build. Why we commit to an outcome and a date instead of a percentage. (2026-08-19) - Can a small business actually afford an AI consultant?
Whether a small business can afford an AI consultant is arithmetic you can run in ten minutes. Here it is, plus when we tell people not to buy. (2026-08-19) - What AI consulting actually costs a small business
What AI consulting costs: our real price bands, the software bill that arrives after the fee, and what a cheap quote is buying you. (2026-08-19) - Real estate AI integrations: the operator's map
The systems real estate and property operators already run, what AI can safely do inside each, how each one lets you in, and the order to connect them. (2026-08-08) - Ontario's 2027 rent cap is 1.9%. Your margin lives in operations now.
Ontario capped 2027 rent increases at 1.9%. When revenue is fixed by law, margin becomes an operations question: where do the hours leak? (2026-08-06) - ChatGPT for landlord automation: a working guide
Automate rent reminders, tenant replies, and maintenance triage with six ChatGPT prompts that work today. Built for small landlords and property managers. (2026-08-02) - Where AI actually lands in a property management operation
AI lands on intake: the phones, maintenance triage, lead capture. It needs a human gate on leasing and money. Automate the intake, never the judgment. (2026-08-02) - Automating Buildium and AppFolio with AI
What the Buildium and AppFolio APIs actually allow, the four automations worth building first, and the writes you should refuse to automate. (2026-07-30) - AI property management tools: a comparison guide
The five buckets of AI tools for property managers, what each one is actually for, what to check in the demo, and how to choose without buying shelfware. (2026-07-22) - What the AI consulting boom gets wrong
Most of the money going into AI help gets spent on decks. What operators actually buy, why generalists miss, and four questions to ask any firm. (2026-07-01) - How to automate property management lead intake
Voice AI, structured extraction, and CRM write-back: the boring stack that turns scattered web, email, and phone leads into qualified rows in your CRM. (2026-06-21) - Your most important AI hire is already inside
76% of CEOs are hiring a chief AI officer. 57% are promoted from inside. Most AI deployments don't fail at the model. They fail without a named owner. (2026-05-19) - Why we named the company "the system"
An Arabic word for order, structure, arrangement. We didn't name the company after a product. We named it after the discipline. (2026-05-15) - Why most AI pilots die after the demo
The pilot worked. The demo got applause. Then nothing. The gap between a model that works and a system that holds, and how to plan for it. (2026-05-04) - The audit-build-operate model
Three phases. Fixed scope. Weekly demos. Why most consulting models are built for change orders, and how a deployment model differs from billable hours. (2026-04-22) - The five questions every audit answers
Where is the time going? The money? What's fragile? What's repeatable? What's the next dollar worth automating? Asked in order, that's the audit. (2026-04-15) - Make.com vs n8n vs custom Claude: pick one
A decision matrix from someone who's shipped all three, written for operators who have to live with the choice for the next two years. (2026-04-08) - Anatomy of a four-week build
The four systems we most often build for a small property operation: intake, maintenance triage, lease renewals, and owner reporting, as one engagement. (2026-03-26)