The best AI assistant for a civil or water engineer is the one that fits the task and lets you check its work. ChatGPT, Claude, Gemini, Microsoft Copilot and Perplexity are all capable general assistants; they differ more in how they connect to your files, tools and sources than in raw ability, and all of them can produce a confident wrong number. Features and plans change often, so check the current version before choosing one for a team.
What engineers actually use AI assistants for
- Reading long documents. Summarising a standard, a tender, a geotechnical report or a model manual, and finding the clause you need.
- Drafting. Report sections, method statements, emails and responses to reviewer comments, which you then edit.
- Checking calculations. Re-deriving a pipe headloss or a detention volume independently, ideally by having the assistant write and run code so the arithmetic is visible.
- Scripting. Python for EPANET, SWMM and GIS workflows, Excel formulas, and batch processing of model results.
- Finding sources. Locating papers, guidance and standards, with links you then open and read.
How the main assistants differ
- ChatGPT (OpenAI): broad general assistant with file uploads and a code execution environment for data analysis, useful for running a calculation rather than trusting typed arithmetic.
- Claude (Anthropic): strong at long documents and careful writing, with file uploads and code execution for analysis; useful for working through a long report or specification.
- Gemini (Google): integrated with Google Workspace, convenient if your team lives in Docs, Sheets and Drive.
- Microsoft Copilot: integrated with Word, Excel, Outlook and Teams under an organisation's Microsoft 365 controls, convenient where client data must stay inside that environment.
- Perplexity: search first, with citations attached to answers; useful for finding sources quickly, which you still have to read.
Most firms end up using two: one inside their office suite for drafting and documents, and one for analysis and code.
Can AI assistants do engineering calculations?
They can, but never accept a number they typed without showing the working. Language models can make arithmetic and unit errors while sounding certain. Ask the assistant to write and run code for the calculation, state every assumption and unit, and check the result against a hand calculation or a model you trust. For anything that goes into a design, the calculation of record still belongs in your own spreadsheet or model.
Rules that keep AI output safe to use
- Client data stays where the contract allows. Use business or enterprise plans with data controls for project information, and check what your contracts say before uploading anything.
- Every number is verified. By code you can read, a hand check or a model.
- Every reference is opened. Assistants can cite sources that do not say what is claimed, or that do not exist.
- The engineer signs, not the assistant. Professional responsibility does not transfer.
Where general assistants stop
A general assistant does not know your network, your model or your telemetry. Questions such as why pressure dropped in zone four last night, or what closing a main would do, need a system connected to your own data and a calibrated hydraulic model. That is the difference between a chat assistant and the operational agents described in AI agents vs chatbots for water utilities, and the wider picture is in AI in water management.
Smart Bhujal builds AI tools connected to a utility's own models and data, delivered as a water decision support system.
