AI tools for consultants: what to use for client work

The short version

Each tool below earns a place for a specific job. You probably won't need all of them.

Prices below are US list prices in dollars, as the vendors showed them on October 17, 2026.

ChatGPT or Claude: pick one to start

For consulting work the two overlap a lot. Both keep a client's files and instructions in a project, remember context across chats, analyze spreadsheets and PDFs, search the web and connect to other tools. Individual plans start at $20 a month for ChatGPT Plus and for Claude Pro billed monthly ($17 a month on Claude's annual plan). Standard seats on ChatGPT Business and Claude Team both cost $25 a month each, or $20 billed annually.

You don't need both, so choose by the work you do most.

• ChatGPT suits you if you record calls on a Mac and want them in the same tool. Record mode is in the macOS app on Plus, Pro and Business, capped at four hours a session. Its projects can also link Google Drive files and Slack channels, depending on your plan.
• Claude suits you if you want each client in its own project. Each project has its own memory, and by default it also uses your memory from chats outside projects. Anthropic's memory guide says to turn off "Use account memory" on the project's page to keep its memory fully separate, so check that setting on each client project. Claude also creates Word, Excel, PowerPoint and PDF files on every plan.

If neither reason applies, try both free plans on the same real task, such as a messy client export and a one-page scope, and keep the one whose drafts you edit least.

Researching a client's market

Your general assistant can already search the web, and both have a longer research mode too. Claude's is on paid plans, and ChatGPT gives Free users a limited amount. Start there.

Perplexity earns a place if you research every week and want each claim tied to a page you can open. Pro is $20 a month, or $17 a month billed annually, and its Projects keep one client's research and files together, private unless you share them.

For example, before a discovery call with a regional HVAC company, you ask which scheduling tools its three nearest competitors mention on their own sites. You open each cited page before you repeat anything to the client, because a citation only tells you where to look.

Skip it if your assistant's research already gives you sources you trust.

Making sense of what the client sends

Spreadsheet exports, PDFs of old process docs and long email threads are a good first job for a general assistant. ChatGPT's data analysis reads .xlsx, .csv and PDF files and runs code to make tables and charts. Claude analyzes uploaded data and can return the result as an Excel file.

For example, a client sends 4,000 rows of lead data and asks why one rep gets most of the quote requests. Ask for counts by rep, source and week, then check a handful of rows yourself before you put a number in front of the client.

No extra tool is needed here. If a job like this keeps coming back, save the instructions in that client's project so next month's export gets the same treatment.

Preparing documents and decks

Both general assistants now make real files. Claude creates Word, Excel, PowerPoint and PDF files on every plan. ChatGPT can create and edit documents, spreadsheets and presentations, with availability that depends on your plan. For a scope document, a status report or a short readout deck, that's usually enough.

A dedicated presentation tool makes sense only if decks are a weekly deliverable. Gamma is one option. It exports to PowerPoint and PDF, and Google Slides by uploading the PowerPoint file. Free exports carry a "Made by Gamma" badge, and Gamma's training policy allows data from individual plans to train its models, while Team and Business are opted out.

Capturing client calls

If you run several client calls a week, a meeting recorder takes over the notes you'd otherwise write afterward. Fathom works with Zoom, Google Meet and Microsoft Teams. The free plan includes unlimited recordings and advanced summaries for your first five calls each month; Premium is $20 a month, or $16 billed annually.

Its consent setting is the useful part for client work. Fathom can email attendees 24 hours before the meeting, and if anyone declines, auto-record turns off for that call. You can let a bot join the call or record without one, though bot-free recording is Mac only and still in beta.

For example, your Thursday status call with a client's operations lead ends with three small changes and one open question. The summary gives you a draft list. You still decide which items become requests and confirm them with the client in writing.

Skip it if you mostly work on a Mac with ChatGPT and want one subscription: ChatGPT's record mode may be enough. Either way, get consent before you record anyone.

Running your own inbox and follow-ups

Lindy and Viktor both pitch themselves as AI coworkers that live where your team already talks and act in the tools you connect. That makes them most useful for your own operations, and worth a careful look before you point them at client work.

Lindy works through Slack, email, iMessage and a Chrome extension for Gmail. Its email drafting and follow-up bumps save replies to your Drafts folder for you to review and send. For example, a client goes quiet after you send a scope on Monday, and on Thursday a polite nudge is waiting in Drafts. Pricing starts at $29.99 per user a month for 3,000 credits. Its approval rules (always allow, require approval, don't offer) apply in shared Slack threads only, not in DMs, web chat or iMessage.

Viktor works in Slack or Microsoft Teams, with a web app too, and is priced per workspace on credits ($25 per 10,000 credits). Its approvals are documented in more detail: an action set to need approval runs only after someone clicks Approve. Viktor's agency page describes connecting the same app once per client and labeling each connection, and weekly client reports reviewed before they go out. The same page answers whether each client gets an automatically isolated context with "Not today."

Choose Lindy if the job is your own inbox. Choose Viktor if a small team wants one assistant across shared tools with approvals on actions. On keeping clients apart, Viktor offers labeled and private connections and asks agencies that need strict separation to talk to its team, and Lindy's docs don't describe it. Check that before client files go in.

Keeping ongoing work for several clients together

A general assistant may be all you need for occasional drafting or a one-off analysis. A project per client in ChatGPT or Claude goes a long way, and both remember context. The job changes when you look after live automations for several clients at once and the same client sends a new request every week.

That's the job AgentHub is built for. You add a client, save a few details about them, and ask questions in that client's conversation. The answer can use those saved details and what you told AgentHub in your earlier conversations about that client, and it never uses another client's conversations. From there it helps turn a client's message into a request draft, makes Word and Excel files, lets the client approve the exact scope from a link without an account, and drafts a client update from the work you recorded.

For example, a client asks whether web quotes can go to reps by zip code. You ask AgentHub about it in that client's conversation, the answer picks up the routing rules you told it about last month, and you turn it into a request for the client to approve.

You still make the calls. AgentHub doesn't change the client's automation or mark work as verified, and nothing reaches the client until you choose to share it. A client update is shared only when you publish it. It is also narrower than the tools above. It has no meeting capture or web research, and it doesn't run or edit automations in n8n, Make or Zapier.

Before you upload client material

Popularity says nothing about how a tool handles client data. Two checks come first: the data settings on the plan you actually pay for, and your client's permission to share the material at all. Contracts often say more about this than people remember. If you keep each client in its own project, also check whether that project's memory is shared with the rest of your account.

• OpenAI may use content from Free, Plus and Pro to train its models unless you turn off "Improve the model for everyone". By default it does not use Business or Enterprise content.
• On Anthropic's Free, Pro and Max plans, check the "Help Improve our AI models" setting before you upload client material. Team and Enterprise content isn't used for training by default.
• Perplexity keeps AI data retention on by default for Free, Pro and Max, and an opt-out covers only data collected after it.
• Fathom, Gamma's individual plans and other tools have their own settings. Read them for the plan you're on, and treat a vendor's security badge as the vendor's claim.

Start with the job that keeps slipping

Pick one general assistant and give it a month of real client work. Note the job that still takes longest. If it's research, calls, your inbox or keeping several clients straight, add the one tool built for it, and check its data settings before the first client file goes in.