AI customer service agents in 2026: what they handle, what they cost, build or buy
What AI customer service agents resolve, what Intercom, Zendesk, HubSpot and others charge per resolution, what building one costs, and the AI disclosure rules.
Field notes
Shipping fast, without shipping fragile.
What we’ve learned about building products in a week, and about taking apps built with AI tools safely to production. Written by the engineer who does the work.
RSS feedLatest
All posts
-
AI receptionists and voice agents in 2026: what they cost, what they handle, and the rules
What an AI receptionist or voice agent costs per call and per minute, what it handles well, how callers react, and the US rules on AI calls and recording.
-
Custom CRM vs HubSpot, Pipedrive or Salesforce: when building your own pays off
What a custom CRM costs next to HubSpot, Pipedrive, Salesforce and Zoho seat prices, when building your own pays off, and what one week of build covers.
-
How to build an AI agent for your business: steps, costs and guardrails
How to build an AI agent in 10 steps: when you need one, which model to use, what it costs to run, and the guardrails that keep it from going wrong.
-
Can you turn a website into an app? What Apple and Google will accept
Converters promise an app in minutes, but Apple rejects repackaged websites. The four routes that work, what each costs, and what passes App Store review.
-
How much does it cost to build an app in 2026?
What an app really costs in 2026: why published ranges disagree, what sets the price, what it costs to run each month, and what fits in one week.
-
Supabase vs Firebase in 2026: which to build on, and what each costs in production
Supabase or Firebase in 2026: databases, security, pricing and lock-in compared, including Firebase SQL Connect and what each really costs in production.
-
What is vibe coding? What it’s good for, and what breaks when real users arrive
What vibe coding means, what it’s good for, what the security research shows, and the checks to run before real users arrive.
-
Supabase RLS not working? A symptom-by-symptom checklist
Empty results, “new row violates row-level security policy”, updates that change nothing, or data anyone can read: each RLS symptom, its cause and its fix.
-
Can you build an MVP in a week? What actually fits
Yes, if it’s one product with one core workflow. What a one-week MVP build can include, what it can’t, and how the scope is set before any code is written.
-
From Lovable demo to production: a checklist
A practical checklist for taking a Lovable (or Bolt, Cursor, Replit) app to production: database access, secrets, auth, payments, monitoring, backups and deploys.
-
The 20 ways AI-built apps fail in production
Apps built with Lovable, Bolt, Cursor or Replit tend to fail in the same ways once real users arrive. The twenty failure modes we check for, and what fixes each one.
-
Why we don’t bill by the hour
Fixed scope and fixed price instead of an hourly rate: why it works better for a first version, what happens when scope changes, and when hourly billing is actually the right choice.
Have something to build?
A free 30-minute call. We’ll tell you honestly what fits in a week, and send a fixed quote afterward.
Prefer to write? Send your requirements