// blog

Build LOG.

Raw writing about vibe coding, building as an introvert, and the Milwaukee AI scene. No SEO filler. Mostly.

pillar

build guides

specific build playbooks
JoshMay 16, 20266 min

An AI-Assisted CP2000 Notice Response Workflow

A CP2000 response is forensic work wrapped around a templated letter. An AI-assisted workflow that does the extraction, cross-referencing and first draft, so the CPA's time goes to review.

read
build guides
JoshMay 15, 20268 min

10 Prompts Every CPA Should Have Saved (With Examples)

Ten prompts for a tax practice, with synthetic examples, source checks and review steps for notice triage, document reminders, research and return-data comparison.

read
build guides
JoshMay 14, 20267 min

The 8 Prompts I Use to Draft Client Emails (And the One I'd Never Skip)

Email drafting is where most professional services time leaks. Eight prompts I use weekly, with the framing that makes the difference, and the meta-prompt that catches every bad draft before send.

read
build guides
JoshMay 14, 20265 min

Copilot in Excel: 8 Workflows That Aren't Just Sum-This-Column

Most teams use Copilot in Excel for trivial autocomplete. Eight real workflows that move the needle on analyst time and decision quality — with exact prompts.

read
build guides
JoshMay 13, 20266 min

Discovery Call Prompts That Don't Sound Like ChatGPT Wrote Them

Six discovery prompts grounded in the prospect’s own words, with a fictional reporting-workflow example and checks for unsupported assumptions, budget and next steps.

read
specific build playbooks
JoshMay 13, 20266 min

Held-Away Account Detection With Plaid + Claude — The Build

Wealth firms often don't know what their clients hold elsewhere. Plaid-connected accounts plus AI pattern recognition can surface rollover opportunities before clients ask. Here's the build.

read
pillar

unfiltered takes

pillar

tool comparisons

tool comparisons
JoshMay 15, 20267 min

Claude Code vs Cursor: Which I Actually Use Day-to-Day

I've shipped real client work in both. They're not the same tool, and the question isn't which one is better. Here's where each one wins and where each one quietly costs you hours.

read
tool comparisons
JoshMay 12, 20266 min

n8n vs Zapier When You're Running a Real Business

Both tools automate. One scales with you. One charges you out of your own workflow. Here's the math that decides which one is right for the size you actually are.

read
tool comparisons
JoshMay 10, 20267 min

ChatGPT vs Claude for Coding: Where Each One Breaks

I've used both for production work and broken both in different ways. Here's the honest map of where ChatGPT wins, where Claude wins, and the one thing that decides it for most people.

read
tool comparisons
JoshMay 8, 20268 min

Make vs n8n vs Zapier: A Three-Way Deathmatch for Small Business

I've stood up all three for paying clients. Each one has a personality and breaks in a specific way. Here's which one I'd pick for which size of business and the one nobody should default to.

read
tool comparisons
JoshMay 5, 20267 min

Perplexity vs Claude vs ChatGPT for Research: Who Actually Cites Real Sources

I researched the same five questions in all three tools. The differences are not what the marketing suggests. Here's where each one wins, where each one fabricates, and the one you should default to for real work.

read
tool comparisons
JoshMay 2, 20268 min

Cursor vs Windsurf vs Cline: The AI IDE Shootout

Three IDEs claiming to be the future of AI-assisted coding. I've built real projects in each one. Here's where each wins, where each disappoints, and which one I'd recommend by skill level.

read
pillar

use cases

use cases
JoshMay 16, 20266 min

Top 10 AI Use Cases for HR (Two Are Copilot Wins for First-Timers)

Ten specific AI use cases for HR teams ranked by leverage. The first two are Microsoft Copilot moves anyone can ship this week. The rest range from quick wins to multi-quarter programs.

read
use cases
JoshMay 15, 20266 min

Top 10 AI Use Cases for Marketing (Two Are Copilot Wins for First-Timers)

Ten AI use cases for marketing teams ordered by leverage. First two are Microsoft Copilot moves the most junior marketer can ship today. The rest scale through demand gen and brand work.

read
use cases
JoshMay 14, 20266 min

Top 10 AI Use Cases for Operations (Two Are Copilot Wins for First-Timers)

Ten AI use cases for ops teams. First two are Microsoft Copilot moves any ops manager can ship in a coffee break. The rest range from operational hygiene to multi-quarter transformation.

read
use cases
JoshMay 13, 20266 min

Top 10 AI Use Cases for Finance (Two Are Copilot Wins for First-Timers)

Ten AI use cases for finance teams ranked by leverage. First two are Microsoft Copilot moves your CFO's analyst can ship by lunch. The rest range from close acceleration to strategic FP&A.

read
use cases
JoshMay 12, 20266 min

Top 10 AI Use Cases for Sales (Two Are Copilot Wins for First-Timers)

Ten AI use cases for sales teams ranked by leverage. First two are Microsoft Copilot moves any SDR can ship in their first hour. The rest range from quick wins to multi-quarter revenue programs.

read
use cases
JoshMay 11, 20266 min

Top 10 AI Use Cases for Customer Support (Two Are Copilot Wins for First-Timers)

Ten AI use cases for customer support. First two are Copilot moves any frontline agent can ship today. The rest range from CSAT improvements to multi-quarter automation programs.

read
pillar

best of 2026

honest tool reviews
JoshMay 16, 20266 min

Microsoft Copilot: What Actually Works, and What Is Theater

An honest map of which Microsoft Copilot features earn their per-seat cost in an enterprise rollout, and which ones get switched off once the novelty wears through.

read
honest tool reviews
JoshMay 8, 20265 min

Claude in Production: What It Is Good At, and What It Costs

Where Claude earns its place in a production feature, where the token bill turns out to be the real constraint, and the cases that belong on a different model.

read
honest tool reviews
JoshMay 6, 20265 min

Cursor for Non-Developers: Who Actually Gets Value

Cursor is sold to developers and adopted by people who are not. Which non-technical roles get real value from it, which bounce off, and what separates the two.

read
honest tool reviews
JoshMay 3, 20265 min

Lovable: What It Builds Well, and Where It Stops

Lovable gets a working app in front of you fast. Which project shapes survive contact with production, which stall, and the point where you outgrow the tool.

read
honest tool reviews
JoshApr 30, 20264 min

Notion AI vs Standalone Tools: When Convenience Wins

Notion AI sits inside the tool millions already use. Convenience is its biggest feature. Here's where it wins, where it loses to standalone tools, and whether to pay for it.

read
honest tool reviews
JoshApr 27, 20265 min

Perplexity Spaces vs ChatGPT Projects vs Claude Projects

Three workspace features solving the same problem from different philosophies. Where each one's design shows through, and which kind of work each is built for.

read
pillar

quick wins

build it in 30 minutes
JoshMay 12, 20265 min

Wire Your Inbox to Notion in 30 Minutes (n8n)

A working pipeline that takes new emails, classifies them, and drops the important ones into a Notion database with summaries. End-to-end build in one sitting.

read
build it in 30 minutes
JoshMay 11, 20265 min

Auto-Summarize Zoom Meetings Into Slack in 30 Minutes

Meeting ends. Slack gets a clean summary with action items, owners, and next steps. No more 'what did we decide on that call?' Build it today.

read
30-min builds
JoshMay 11, 20266 min

Copilot Studio for Non-Developers — Ship Your First Agent in 30 Minutes

Microsoft Copilot Studio lets non-developers build conversational agents that work inside Teams, SharePoint, or web. Here's the 30-minute build path from zero to a working agent.

read
build it in 30 minutes
JoshMay 9, 20265 min

Build a Lead-Qualification Webhook in 30 Minutes

Inbound form fires. Lead gets scored. Sales gets a Slack ping with the score and a one-paragraph summary. Cold leads get an auto-decline. Build the whole thing today.

read
build it in 30 minutes
JoshMay 7, 20265 min

Build a Daily KPI Digest Email in 30 Minutes (No Code)

Morning coffee. Email arrives. Numbers you care about, with day-over-day deltas and a one-line take. Build the whole thing from scratch without writing code.

read
build it in 30 minutes
JoshMay 5, 20265 min

Auto-Draft Proposals From a Google Doc Template in 30 Minutes

Fill out a short brief. Get a polished proposal draft generated from your template, ready to edit and send. Save 45 minutes per proposal. Ship it today.

read
pillar

state of ai 2026

state of ai 2026
JoshSep 14, 20268 min

The 95% AI Failure Stat: What the Study Actually Says

Every article about AI disappointment cites one number from one MIT working paper. I read it. The 95 percent is real, it does not mean what it is used to mean, and the same study found general-purpose AI tools succeeding above 80 percent.

read
state of ai 2026
JoshSep 14, 20269 min

AI Ran the Attack. What That Report Actually Found

Anthropic published a threat report describing AI doing most of the work in real intrusions, with humans supervising. The coverage has mangled it in both directions. Here is what the report says, what it does not, and what changes for the people defending a bank or a firm.

read
state of ai 2026
JoshSep 14, 20267 min

How Much Time Does AI Actually Save? The Numbers, Reconciled

Published estimates of the time AI saves run from about one hour a week to seven. They are not contradicting each other so much as measuring different things with different methods. Here is what produced each number, and the rework finding that eats a chunk of all of them.

read
state of ai 2026
JoshSep 14, 20268 min

If the AI Bubble Bursts, What Happens to Your Vendors?

The bubble argument is usually told from the investor's seat. Almost nobody tells it from the seat of a firm that has put AI into its workflows and would like it to still be there in three years. The arithmetic that drives the worry, and what to actually do about it.

read
specific build playbooks
JoshApr 21, 20266 min

Anthropic Prompt Caching: How the Pricing Works and What It Saves

Prompt caching bills repeated input at a fraction of the normal price, and what it saves on a real bill depends on how much of each prompt repeats and how often. How the pricing works, what breaks the cache, and how to estimate savings from your own traffic.

read
pillar

for veteran owners

for veteran owners
JoshMay 16, 20265 min

Why Combat Veterans Are Unusually Good at Managing AI Agents

Mission planning, after-action reviews, command structure thinking — the muscle veterans built in service maps directly to how AI agents need to be supervised. Here's why a veteran business owner has an advantage that most civilian founders don't.

read
for veteran owners
JoshMay 15, 20265 min

The OODA Loop Mapped Onto AI Agent Design

Observe, Orient, Decide, Act. The framework that keeps fighter pilots and infantry leaders alive is the cleanest framework for designing AI agents that don't get stuck. Here's the mapping.

read
for veteran owners
JoshMay 14, 20265 min

From Mission Planning to AI Workflow Orchestration

The same brain that ran a route reconnaissance plan can orchestrate an AI workflow. Here's the direct mapping from military mission planning to AI orchestration design.

read
for veteran owners
JoshMay 12, 20265 min

SOP Discipline: Why Veterans Deploy AI More Reliably Than Civilians

Standard Operating Procedures are boring. They're also the reason military units work and civilian teams thrash. Here's how SOP discipline translates to deploying AI without setting your business on fire.

read
for veteran owners
JoshMay 10, 20265 min

Squad-Level Thinking Applied to Multi-Agent Systems

Use role and handoff planning to evaluate a multi-agent workflow. Compare it with a simpler baseline and test source evidence, retries, permissions and recovery.

read
for veteran owners
JoshMay 8, 20265 min

After-Action Review for AI Implementations

Review an AI workflow using expected results, observed evidence, causes and verified actions. A fictional example shows why the denominator changes the conclusion.

read
// ready to talk?

Reading about it is the easy part.
Now let's ship it.

Tell us what's broken and we'll tell you straight up whether we can help. No pitch deck. No discovery gauntlet.

Browse every post