Hank vs ChatGPT, Claude, Perplexity, and Gemini: Why General-Purpose AI Fails Contractors

"Why can't I just use ChatGPT?"

It's the question every contractor asks when they see a purpose-built construction AI for the first time. Fair question. General-purpose AI has become genuinely impressive in 2025–2026. ChatGPT can write code. Claude can summarize a 200-page contract. Perplexity can research a supplier faster than a human. Gemini can analyze a photo. Why pay for a specialized tool when the general ones are so capable?

Here's the honest answer, from a contractor who spent six months trying to run a jobsite on ChatGPT, Claude, Perplexity, and Gemini before giving up and building a purpose-built system.

General-purpose AI fails contractors for one reason: it doesn't know your business. Everything downstream of that fact — the wrong numbers, the made-up margins, the copy-pasted scope that doesn't match your estimate, the endless context switching — comes from the same root problem.

This isn't a knock on the general models. They're brilliant. But being brilliant isn't enough. A contractor needs AI that's grounded in the specific data of their specific jobs, and no general-purpose tool can provide that out of the box.

Here's the comparison in detail, tool by tool.

ChatGPT (OpenAI)

What it's good at for contractors

• Writing boilerplate content (generic scope language, cover letters, email templates)
• Explaining construction concepts you haven't worked with before
• Converting between units, doing quick math
• Generating formulas for spreadsheets
• Summarizing long documents you paste in
• Image analysis on photos (ChatGPT 4o and newer)

Where it fails for contractors

It has no access to your data. If you ask ChatGPT "what's my average margin on HVAC installs this year?" the answer is literal fiction. It makes up a number that sounds plausible. There is no way to ask ChatGPT about your actual jobs because it cannot see them.

Context evaporates between sessions. You paste in project details in one chat. Next week you come back, start a new chat, and have to paste everything in again. There's no persistent project memory. No way to pin a client or a job. Every conversation starts from zero.

It gives construction advice that's wrong in specific, dangerous ways. Ask ChatGPT for labor production rates for hanging drywall and you get numbers from a textbook. Those numbers are averages of averages — useless for bidding your crew's work on this job. Use them in your estimate and you will lose money.

Actions are text-only. You can't ask ChatGPT to create a task, advance a pipeline deal, or update an estimate line. It will describe what you should do. It can't actually do it.

Image analysis is not takeoff. Yes, ChatGPT 4o can look at a photo of a blueprint. No, it cannot reliably measure the areas on that blueprint to within the accuracy a bid requires. If you try to use it for takeoff, you will get crushed.

The honest verdict

ChatGPT is a fantastic general writing and research assistant. For a contractor, it's a drafting tool — boilerplate, quick research, generic emails. It is not, and will never be without fundamental changes, a tool that can answer questions about your actual business or make decisions on your data.

Claude (Anthropic)

What it's good at for contractors

• Reading and summarizing contracts and long specs (Claude's long-context ability is the strongest of any general model)
• Drafting nuanced client communications (tone control is excellent)
• Explaining reasoning step by step
• Handling multi-step analysis in one prompt
• Image analysis on photos and drawings

Where it fails for contractors

Same root problem as ChatGPT: no access to your data. Claude cannot see your estimates, your projects, your cost database, or your client records. When you ask it about your business, it has nothing to ground its answers in except what you pasted in this session.

The conversation context resets. Claude has an enormous context window — you can paste in a huge project — but it's still per-conversation. Tomorrow's chat doesn't remember yesterday's scope.

No write access to real systems. Even if Claude reads and summarizes a proposal perfectly, it can't store the result, update an estimate line, or create a project. You take the output, switch tabs, and re-type it into your real system. That's half an hour of lost time per task.

Pricing scales per conversation length. Every time you paste in a long document to give context, you're paying for the input tokens. Running a construction business on Claude adds up fast if you're serious about using it.

The honest verdict

Claude is the best general-purpose AI for deep contract reading, nuanced drafting, and reasoning work. For a contractor, it's a one-off analysis tool — read a spec, draft a tough email, review a contract. Useful and sometimes essential, but it cannot run your business. The underlying model (Claude from Anthropic) is actually what powers purpose-built contractor AIs like Hank — the difference is that Hank is connected to your real data and wrapped in a toolkit designed for construction work.

Perplexity (and Perplexity Comet)

What it's good at for contractors

• Research that requires fresh web data (supplier prices, material availability, code updates)
• Fact-checking with citations
• Finding answers that need both reasoning and up-to-date information
• Comparing products or suppliers
• With Comet (the browser agent), basic automation across web pages

Where it fails for contractors

It's a research tool, not an operating system. Perplexity is designed to answer questions by searching the web. It doesn't hold your data. It doesn't know your business. It can tell you that a 2x4 costs $5.89 at Home Depot today, but it can't tell you what you paid last time, what your margin was, or which of your projects is using that size lumber right now.

Comet (the browser agent) is impressive but fragile. Agentic web browsing is powerful — it can fill forms, scrape data, compare prices across suppliers. But it's slow, unreliable on complex sites, and has no durable memory of what it did for you last week. Not a fit for repeatable contractor workflows.

No connection to your real data. Same core problem. Perplexity's strength is the open web; your business data lives in a closed database that Perplexity has never seen.

The honest verdict

Perplexity is the best research tool on the market. For a contractor, it's what you open when you need to research a supplier, check a code update, or compare materials. It is not a tool that answers questions about your jobs, and Comet's agentic abilities, while impressive, don't replace a purpose-built business system.

Google Gemini

What it's good at for contractors

• Integration with Google Workspace (Docs, Sheets, Calendar, Gmail)
• Multimodal analysis (text, image, video, audio in one model)
• Summarizing email threads inside Gmail
• Long-context handling similar to Claude

Where it fails for contractors

Integration is with Google, not with your contractor stack. Gemini is excellent if your entire business runs on Google Docs and Gmail. Most contractors don't. Your estimates are in estimating software. Your projects are in project management. Your costs are in a cost database. Your invoices are in QuickBooks or equivalent. Gemini sees none of that.

It has the same context-resets-per-session problem as Claude and ChatGPT. No persistent project memory, no pinned context, no job-level continuity.

Construction-specific reasoning is generic. Gemini knows what a "footing" is. It doesn't know what your specific crew's production rate on footings is, what your current foundation project is, or what your concrete supplier charged you last month.

Workspace Gemini can write to your Google Docs and Sheets, which means it has action capability there — but not in the tools that actually run construction businesses. Creating a Google Sheet with a bid summary doesn't help if your bid lives in your estimating system.

The honest verdict

Gemini is strong if your business runs on Google Workspace. For a contractor who lives in Docs and Sheets, it's useful for drafting and summarizing. It's not a tool that knows your construction business — it knows your Google data, which is a subset of what matters.

The pattern: all general-purpose AI shares the same core weakness

Every tool above has the same problem. It can reason, write, research, and analyze. It cannot see your construction business. It has no access to your estimates, projects, costs, clients, or invoices. So every answer it gives about your business is either generic or made up.

The fix is not a smarter model. The fix is a model that's connected to your real data.

What purpose-built contractor AI does differently

Hank — the AI assistant built into Contractor Co-Pilot — uses the same underlying technology as Claude. Same brilliance, same reasoning quality. The difference is that Hank is connected to your actual business database through roughly 40 purpose-built tools. Ask Hank "what's my average margin on HVAC installs this year?" and the answer is:

1. Hank calls the query_estimates tool with a filter for HVAC work this year.
2. The tool returns real rows from your actual database.
3. Hank calculates the margin from those real numbers.
4. Hank gives you the answer, along with the underlying jobs it used.

That's it. No hallucination, no made-up numbers, no generic averages. The answer is either "your real margin is X" or "you don't have enough data yet, here's what you have." Both are useful. Neither is fiction.

The five things Hank does that general-purpose AI can't

1. Hank has read access to your live business data. Estimates, projects, clients, cost database, calendar, email, invoices. Every query returns real numbers from real records.

2. Hank has write access with explicit confirmation on every action. Create a task, advance a pipeline deal, update a budget, draft an invoice. Hank shows you what it's about to do; you confirm; it runs. You can never get a silent write.

3. Hank persists context across conversations. Pin a project or client once and the context stays. Come back tomorrow and Hank still knows which job you're working on. No re-pasting, no re-explaining.

4. Hank is auditable. Every tool call is logged with the user, timestamp, parameters, and result. You can see exactly what Hank did and why. A general-purpose AI is a black box; Hank is a glass box.

5. Hank speaks contractor. Scope, margin, variance, labor burden, waste factor, retainage, buyout, punch list — Hank understands these in the context of your actual jobs. General AI knows the vocabulary. Hank knows how the vocabulary applies to your business.

When should you use general-purpose AI vs Hank?

Here's the honest split. Use the right tool for the job.

Use general-purpose AI (ChatGPT, Claude, Perplexity, Gemini) when:

• You need to research something on the open web (Perplexity is best)
• You're drafting a nuanced client email on a one-off situation (Claude or ChatGPT)
• You're reviewing a complex contract or spec and need a second set of eyes (Claude is best for long-context reading)
• You're learning a new trade or technique you haven't worked with before
• You need generic formulas, quick math, or unit conversions
• You're brainstorming marketing content or blog posts

Use Hank (or equivalent purpose-built contractor AI) when:

• You need an answer about YOUR business ("what's my margin on X?")
• You need to audit YOUR estimate before it ships
• You're matching takeoff lines to YOUR cost database
• You need to draft a client email WITH your project context
• You want to advance a pipeline deal, create a task, or update a job
• You're trying to find a specific job, client, or invoice in YOUR records
• You need to compare a current bid to YOUR historical jobs
• You want AI to flag what's missing from YOUR estimate based on YOUR history

Neither is "better." They serve different jobs.

A contractor in 2026 probably uses both. Perplexity to research a supplier. Claude to review a tough contract. ChatGPT to draft generic copy. Hank to actually run the business. The mistake is thinking one of them replaces the others. They don't. They each serve a different purpose.

The other mistake is thinking general-purpose AI will someday replace purpose-built contractor AI. It won't. Not because general AI isn't good enough — it's plenty good enough — but because the value of a contractor AI is access to your data, and general-purpose tools can't get that access without being connected to your business systems. The moment a general AI is connected to your real estimates, projects, and costs, it stops being general — it becomes, functionally, a purpose-built contractor AI.

The contractor who tried running on ChatGPT for six months

I know a landscaper in Ohio who spent six months trying to run his business on ChatGPT. He'd open a new chat every morning, paste in that day's projects, ask for help drafting bids, and copy the results into his estimating spreadsheet. Every evening he'd ask ChatGPT to summarize what happened on each job that day, then paste the summary into his project tracker.

It worked. Kind of. It saved him maybe an hour a day on drafting. But after six months he added it up: he was spending 30 minutes every morning re-pasting context, copying things between tools, and re-explaining which job he meant. The net time savings was 30 minutes. The net stress savings was zero. He'd traded one type of work for another.

When he switched to Contractor Co-Pilot and Hank, the context management disappeared. Hank knew every project because every project lived in the same database. He stopped copy-pasting. He stopped re-explaining. The AI was the same underlying technology — Anthropic's Claude — but now it was connected to his actual jobs.

That's the difference. That's what "purpose-built contractor AI" actually means. It's not a smarter model. It's a model that finally has the context it needs to be useful.

Related reading

• The AI operating system for contractors — how AI fits into the whole business, not just one step
• AI for contractors in 2026: what's real vs. hype — our honest guide to the category
• Meet Hank — the contractor AI assistant — the tool this post is about
• AI estimating software — the specific workflow where Hank earns its keep
• Contractor Co-Pilot vs spreadsheets — the other comparison worth reading