You can use Claude to read construction drawings and draft scopes of work. In our example, one pass over a 1,200-page project costs about $9.40 on Claude's API. API charges are only part of the cost. A production workflow also needs document handling, bid comparison, revision tracking, and software upkeep. For occasional questions, Claude alone may be cheapest. For teams running many projects, a platform like Consight replaces that build.
Consight publishes this comparison and sells one of the two products in it. Claude is described from Anthropic's public documentation, and the method is at the end of the page.
General contractors are asking whether they can build their own AI estimating workflows on Anthropic's API, now that Claude Skills and stronger models exist. The economics look good at first, because Claude can process construction documents at a low cost per request.
The better question is the total cost. What does it take to turn hundreds or thousands of pages into accurate scopes of work and leveled subcontractor bids?
Can Claude AI read construction drawings?
Yes. Claude reads text, pictures, charts, and tables in PDFs, including drawings, specs, schedules, and construction notes.
Estimators can ask Claude to:
- Name the roofing system the drawings specify.
- Identify the door hardware requirements.
- Summarize the Division 09 specifications.
- Draft a preliminary scope of work for the drywall subcontractor.
- Find discrepancies between two specification sections.
Claude Skills make these tasks more repeatable. A Skill packages instructions, metadata, and optional resources such as scripts and templates. Anthropic's engineering team compares a Skill to an onboarding guide for a new hire.
For an estimator researching one detail, that can be valuable. Construction estimating involves more than answering questions about drawings, though.
A general contractor must find scope across many trades and organize bid packages. The team also tracks clarifications and addenda, collects proposals, compares exclusions, and keeps records of what it reviewed and awarded. Those jobs take more than an AI model.
How much does Claude's API cost for construction plan reading?
Anthropic charges API customers mainly by input and output tokens. As of November 2026, Anthropic's published standard rates for Claude Sonnet 5.5 are:
| API usage | Price |
|---|---|
| Input tokens | $2 per million tokens |
| Output tokens | $10 per million tokens |
Claude bills a PDF page as both text and an image. Anthropic says a page typically uses 1,500 to 3,000 text tokens, and the page image adds more.
Example: a 1,200-page commercial construction project
Consider a project with:
- 200 architectural, structural, and MEP drawing sheets
- 1,000 pages of specifications
- 1,200 total document pages
Assume each page uses an average of 3,500 input tokens for its text and visuals. That is a low-end assumption, and a pass could cost about twice as much on image-heavy sets. Assume each analysis pass generates 100,000 output tokens.
| Analysis activity | Estimated Sonnet API cost |
|---|---|
| One complete document analysis | $9.40 |
| Three complete analysis passes | $28.20 |
| Six complete analysis passes | $56.40 |
| Ten complete analysis passes | $94.00 |
These are modeled estimates, not benchmark results. Actual costs depend on document complexity, model choice, context management, and how an application processes the documents.
A well-built Claude application can cut these costs. It can avoid rereading every page on every request, and Anthropic discounts cached input. Its Batch API also takes 50 percent off work that can run asynchronously, so six passes would drop from $56.40 to $28.20.
An application must also split the job. Claude accepts up to 600 PDF pages per request, and our example's 4.2 million input tokens exceed one 1 million token context window.
The takeaway: reading plans with Claude can be inexpensive. Completing and maintaining a full preconstruction workflow is a different calculation.
What are the hidden costs of building estimating workflows with Claude Skills?
Claude Skills automate specialized tasks, but they do not give you a complete estimating platform. To reach production, a contractor or development team must build the system around Claude.
1. Document organization and retrieval
Construction projects often hold hundreds of drawing sheets, specification sections, addenda, and revisions.
An estimating system must find the right document version and the relevant information. It must also keep references and pull the right context for each task. Without that structure, estimators may spend real time finding, uploading, and organizing project files.
2. Scope of work generation
A preliminary scope of work is only the start.
Commercial contractors organize scope by trade, review coverage, find missing requirements, and resolve overlaps. They also keep bid packages consistent. A custom Claude solution needs prompts, data structures, and validation to do that reliably.
3. Bid leveling and subcontractor comparisons
After subcontractors bid, estimators compare each proposal against the project requirements. That means reviewing exclusions, allowances, alternates, qualifications, and possible scope gaps.
An AI model can help analyze this information. A contractor still needs a structured system to compare bidders, keep decisions, and produce bid leveling reports.
4. Revisions and addenda
Preconstruction documents rarely stay the same. Addenda, revised drawings, bidder clarifications, and updated proposals can force the team to revisit scopes and comparisons.
A custom AI solution must handle those changes without losing reviewed work or forcing duplicate work.
5. Engineering and maintenance
Even with an affordable API, a custom solution brings ongoing software costs. Those include development, hosting, integrations, permissions, document storage, monitoring, testing, and support.
For most general contractors, running an internal AI software platform is not the core business.
Claude AI vs. Consight: feature comparison
Consight is an AI preconstruction platform built for commercial general contractors. It connects construction document analysis with scope development and subcontractor bid management, so estimators do not assemble separate AI tools.
| Capability | Claude API + Skills | Consight |
|---|---|---|
| Read construction drawings and specifications | Yes | Yes |
| Answer questions about construction documents | Yes | Yes, through Connie |
| Generate trade scopes of work | Drafts from a prompt; repeatable output needs a configured workflow | Built in |
| Organize bid packages | Requires custom implementation | Built in |
| Link scope requirements to source documents | Requires custom implementation | Supported workflow |
| Compare subcontractor bids | Compares proposals you upload; a repeatable system needs custom work | Built in |
| Identify exclusions and scope gaps | Requires workflow and validation | Built in |
| Manage bid leveling | Requires custom application logic | Built in |
| Produce structured preconstruction reports | Pre-built document Skills help; your report format needs setup | Built in |
| Support shared project workflows | Requires application infrastructure | Built in |
Claude can perform these tasks with the right application around it. Consight is that construction-specific application and workflow, already built for estimating and preconstruction.
Consight has drawbacks too. You pay a subscription in slow months, you work within its workflows rather than your own design, and team pricing is not published. A custom Claude build gives you full control of all three, and you own the upkeep.
Comparing the real cost: Claude API vs. Consight
Consider a commercial general contractor that processes 10 projects a month. Using the 1,200-page example, six full analysis passes per project come to about $564 a month in Claude Sonnet API charges.
Now assume the custom workflow adds 30 minutes of manual formatting, review, and document organization per project. At an illustrative fully loaded estimator cost of $75 an hour, that adds $375 a month.
| Monthly cost component | Illustrative cost |
|---|---|
| Claude API usage | $564 |
| Additional estimator labor | $375 |
| Combined monthly cost | $939 |
| Custom engineering and maintenance | Not included |
This is a scenario, not a claim that every Claude implementation costs $939 a month. Optimized workflows may use fewer tokens, and batch pricing alone would cut the API line to $282. Manual labor varies, and Consight's output needs estimator review too.
Consight Pro is one plan, priced by quote for your whole organization. Every member is included, and each month includes an AI budget; additional usage is billed separately.
A quote prices Consight for your company and your volume, so this page cannot state Consight's cost for this scenario. Get a quote for your own volume.
This is not a like-for-like comparison. It shows why contractors should weigh the total cost of usable estimating work, rather than set a subscription price against raw token charges.
Why does purpose-built AI matter for preconstruction?
Purpose-built AI matters because estimating is a specialized workflow, not a document summary. An estimator needs to know exactly what belongs in a subcontractor's price, find missing scope, and make reliable buying decisions.
That takes continuity between these steps:
Plans and specifications → Scope generation → Bid packages → Subcontractor proposals → Bid leveling → Award decisions
In disconnected tools, estimators may have to move information by hand, recheck earlier work, and keep separate records.
Consight brings these steps into one shared platform. Contractors can standardize how their teams work across projects, and people still review the results and make the decisions.
Is Claude AI enough for construction estimating?
Claude can be enough for targeted questions, document summaries, and exploratory analysis. For an estimator who occasionally reviews a spec or drafts a scope, using Claude directly may be the cheapest option. This page prices the API. Our guide to AI for construction estimating covers Claude and ChatGPT plans, and what chat tools do well for estimators.
Commercial general contractors running many projects, trades, subcontractors, and estimators face a bigger decision than the cost of AI inference. A production workflow needs organization, traceability, collaboration, document versioning, and structured outputs.
A team can build those capabilities around Claude, at added cost. A contractor that would rather estimate than develop software may get more from a purpose-built platform.
The bottom line
Claude can read construction drawings. The real question is whether a general contractor wants to build and run its own AI estimating system. The other path is a platform that already connects document analysis to the rest of preconstruction.
For occasional document questions, Claude may be an affordable answer. For complex preconstruction workflows, the better measure is the total cost of accurate, reviewable, usable estimating work.
That is the problem Consight is built to solve. See how plan reading, scope generation, and bid leveling work together.
How we compared
We reviewed Anthropic's public API pricing, PDF support, and Agent Skills documentation in November 2026, and Consight's own product and pricing pages. Cost tables are models built on stated assumptions, not measured Claude workloads. The token counts, passes, labor time, and $75 hourly rate are illustrative. Prices change, so check each vendor's current pricing before you decide.
Where Consight fits
Consight is the preconstruction workspace built around this kind of work—from understanding project documents and building scopes to collecting bids, leveling proposals, reviewing risk, and documenting the final award.
Put it into practice on your next project.
Talk to sales