The AI Tool Stack a One-Person Consultancy Actually Needs

Most one-person consultancies need four things: one paid chat assistant (ChatGPT Plus or Claude Pro at $20 a month), the AI built into their email and documents suite, and a meeting note-taker, plus either a research tool or one automation platform. That comes to roughly $30 to $100 a month. Add anything else only when a specific weekly task demands it.

Two things trip solo consultants up. First, seat minimums: ChatGPT Business and Claude Team both need at least two seats, so most solo practices sit on individual plans and must switch off the model-training setting themselves. Second, overlap. Paying for three chat assistants, each used a little, is the most common waste I see, and it spreads your prompts, files and client context across tools that never talk to each other. For the general question of how many tools any small business needs, see how many AI tools a small business actually needs; what follows is specific to working alone.

Follow me on Instagram@sagnikteaches

1. One chat assistant you use every working day

What it is. The general-purpose assistant you draft, summarise, analyse and think with: ChatGPT Plus ($20 a month), Claude Pro ($20 monthly or $17 a month billed annually) or Gemini through Google AI Pro ($19.99). Any of them handles proposal drafts, meeting follow-ups, reading a client's 40-page strategy document and tidying a spreadsheet of survey responses.

Connect on LinkedInSagnik Bhattacharya

How to choose. Run the same five tasks from your real week through two free versions for a few days: a proposal section, a summary of a long document, a client email, a piece of analysis, a slide outline. Pick the one whose drafts you edit least. Don't choose on benchmark tables; choose on your own tasks.

Subscribe on YouTube@codingliquids

What it looks like in practice. Take an illustrative solo consultant who advises IT support firms on service pricing. They keep one Project per client, holding the engagement letter, the latest pricing model and the notes from each call, with standing instructions at the top. Monday morning, one prompt:

Using the files in this project, summarise where this engagement stands
in five bullet points: what we agreed, what I've delivered, what's
overdue on either side, the next deliverable and its date, and any
decision the client still owes me.

An illustrative reply:

- Agreed: review of three support tiers and a new per-device price.
- Delivered: tier analysis (12 Sept), draft price sheet (19 Sept).
- Overdue: client's ticket-volume export, requested 15 Sept.
- Next: final pricing recommendation, due 3 October.
- Decision owed: whether after-hours cover moves into the top tier.

The one thing to check is the dates. Assistants sometimes pull a date from a draft file rather than the sent version, so glance at your sent folder before quoting any of them back to the client.

Watch-outs. Don't build a custom GPT: OpenAI is retiring custom GPTs, and existing ones stop running on 11 December 2026, with plugins and skills as OpenAI's replacement. Projects (ChatGPT or Claude) and Gems (which become skills from November 2026) in Gemini are the right home for per-client context. And turn off the model-training option in your privacy settings before the first client file goes in; stopping AI tools training on your business data shows where it lives in each tool.

2. The AI already inside your email and documents

What it is. If your practice runs on Microsoft 365 or Google Workspace, some AI is already on your bill. Microsoft 365 business plans include Copilot Chat, web-grounded, at no extra cost; Business Standard is $14 a user a month on annual billing. Google Workspace now builds Gemini into its plans: Business Starter (about $7) gets Gemini in Gmail and the Gemini app, and Business Standard (about $14) adds Gemini across Docs, Sheets, Slides, Meet and Drive. Confirm the current figures on your account's pricing page, because Workspace prices vary by billing option.

When it's enough. A consultant on Workspace Business Standard who mainly writes documents and emails may find Gemini covers most of what a separate assistant would do, with the advantage that it can see the files already in Drive. A consultant whose work is heavy analysis of client documents usually still wants item 1.

The paid upgrade. Microsoft 365 Copilot Business, which puts Copilot inside Word, Excel, Outlook and Teams, is $21 a user a month on annual billing, with a promotional $18 on annual plans until 31 December 2026. If you need both, the Business Standard plus Copilot bundle at $23.50 a month (annual) costs less than buying the two separately.

What it looks like in practice. A solo marketing consultant who works with web design studios receives a long email thread from a client's founder, three people deep, about a delayed launch. Gemini in Gmail summarises it into what was decided, what is disputed and what the founder is asking the consultant to do. The consultant then writes the reply themselves, because the tone of that email matters more than its speed.

Watch-outs. Microsoft 365 Premium ($19.99 a month) is the consumer replacement for the retired Copilot Pro. It is sold for personal use, so a consultancy handling client data is better served by a business plan with business data terms and admin controls, even with one user.

3. A meeting note-taker for client calls

What it is. A tool that records or listens to calls and produces a transcript, summary and actions. You may already have one: meeting summaries are included in paid Zoom Workplace plans, and Google Meet's "Take notes for me" works for hosts on eligible Workspace plans. Standalone options include Otter.ai (Pro at $8.33 per user monthly on an annual plan, or $16.99 paying month to month) and Fathom (a free plan that records and transcribes without a cap, with fuller AI summaries on Premium at $16 a month annually or $20 monthly).

Why it earns its place. For a consultant, the transcript of a discovery call is raw material for the proposal: the client's words for their problem, the figures they mentioned, the objections they raised. Pasting a transcript into your assistant with "draft the 'understanding of your situation' section of a proposal in the client's own words" saves real time and makes the proposal sound as if you listened, because you did.

A realistic mistake. An illustrative consultant working with a PR consultancy connects a note-taker to their calendar with default settings. It joins a call the client is hosting, announces itself, and emails the summary to all six attendees, including the client's own client, who was on the call. Nothing sensitive was said, but it was awkward. Turn off auto-join for meetings you don't host and check who receives summaries before the first call. The note-taker comparison covers these settings tool by tool.

Watch-outs. Ask permission at the start of every recorded call, every time, and have a no-recording fallback: a three-line voice memo into your assistant straight after the call.

4. A research tool, only if research is part of what you sell

What it is. Perplexity Pro ($20 a month) answers with citations by design. Each of the main chat assistants also has a research mode that searches widely and returns a report with its sources: ChatGPT's deep research (capped monthly, with the cap set by your plan), Claude's Research on paid plans, and Gemini's Deep Research.

When to pay separately. If market research, competitor scans or sector briefings are deliverables you charge for, a dedicated research tool plus your assistant can be worth $40 a month between them. If research is only discovery-call prep, the research mode in the assistant you already pay for is usually enough. Perplexity against ChatGPT for business research walks through the difference on real tasks.

What it looks like in practice. A solo consultant producing a quarterly market note for translation agencies asks Perplexity for "changes in the last 90 days affecting small translation agencies' pricing and demand, with dates and sources". The useful output isn't the summary; it's the list of sources, which the consultant reads, discards half of, and cites properly in their own note. The rule doesn't change with the tool: nothing goes into a client deliverable until you have opened the source yourself.

5. One automation platform, once you have a repeat job

What it is. Zapier or Make, which connect your apps so that something happening in one triggers actions in others, optionally with an AI step in between. Zapier's Professional plan costs $19.99 a month billed annually ($29.99 monthly) for 750 tasks, where each successful action step counts as a task and triggers and filters don't. Make starts free for 1,000 credits a month, with paid plans from about $9 a month.

When it's worth it. Only when you can name a job you do the same way more than about ten times a month. Common ones for consultants: a new enquiry form that creates a contact, files the enquiry and drafts a reply for you to check; a finished call whose summary is saved to the right client folder; an invoice paid that triggers a thank-you and a feedback request.

The maths, filled in. An illustrative solo consultant gets 40 enquiries a month through a website form. Their Zap has a trigger (free), a filter to drop spam (free), then three actions: create a CRM contact, run an AI by Zapier step to draft a reply (1, 3 or 5 tasks depending on the model tier chosen; say 3), and save the draft to Gmail. That is 1 + 3 + 1 = 5 tasks per enquiry, or 200 a month, comfortably inside Professional's 750. The same consultant adding a second Zap for post-call filing, at 60 calls a month and two tasks each, would still be under 350.

Watch-outs. Automations nobody looks at break quietly. Put a monthly reminder in your calendar to open each Zap's history, and keep the AI step drafting rather than sending. For freelancers' specific admin automations, invoices, proposals and chasers with AI has worked set-ups.

How the layers share one engagement

Seeing the stack across a single piece of work shows where each layer earns its money and where the gaps are. An illustrative engagement: a solo operations consultant is hired by a six-person translation agency to fix how projects are quoted and scheduled.

1. Enquiry. The agency's owner fills in the website form. The automation files the enquiry, creates a contact and puts a draft reply in the consultant's inbox. The consultant edits the draft and books a call. Two minutes instead of ten, and the enquiry never sits unanswered over a weekend.
2. Before the call. The assistant's research mode produces a short brief on the agency: languages offered, sectors named on the site, a recent vacancy for a project coordinator. The consultant checks three facts and writes two questions.
3. Discovery call. The note-taker records, with permission. The transcript captures the owner saying "we lose a day every week chasing freelancers for availability", which becomes the first line of the proposal's problem statement.
4. Proposal. In the client's Project, the assistant drafts the situation section from the transcript and the scope from the consultant's standard modules. The consultant rewrites the price and the approach section by hand, because those are what the client is really buying.
5. Delivery. The office-suite AI helps with the working files: Copilot in Excel or Gemini in Sheets cleaning an export of 300 past quotes, and summarising which language pairs run late most often. The consultant checks the counts against the raw data before any finding goes into a slide.
6. Reporting and close. Weekly updates are drafted from the Project's notes. The final report is written by the consultant, with the assistant used to tighten wording and check that every recommendation has an owner and a date.

Notice what never happened: no specialist AI tool was needed at any stage. Notice too where the human work concentrated: the price, the approach, the checks on numbers and the final report. That is the shape of a well-sized solo stack. The tools take the typing and the first drafts; the judgement stays where the client expects it.

What a solo consultant can safely skip

• A second and third chat assistant. Keep one as your main tool. If you want a second opinion occasionally, the free tier of another is enough.
• Standalone AI writing tools. They mostly wrap the same models you get in item 1, with templates you'll outgrow. Your own prompts in a Project do the same job.
• AI proposal software. Worth a look for firms sending dozens of proposals a month. For a solo practice sending four, a good template and your assistant win.
• Custom GPTs. Being retired, as above. Don't build anything new on them.
• A consumer AI plan on top of Workspace. If you pay for Workspace Business Standard, Google AI Pro mostly duplicates Gemini you already have.
• API accounts, unless you're building an automation that needs one.

Three stacks, priced out

List prices in USD, per month, using annual billing where it is cheaper. Check each vendor's page before buying; promotions and plan names change.

The heavy stack has deliberate overlap, Copilot and ChatGPT, for a consultant whose deliverables live in Word and Excel but whose thinking happens in a chat window. Most people should start lean and move up one layer at a time, when a task in the week proves the need.

A before and after makes the point. An illustrative solo consultant audits their card statement and finds ChatGPT Plus ($20), Claude Pro ($20), Perplexity Pro ($20), Google AI Pro ($19.99) on top of Workspace Business Standard, Otter Pro on monthly billing ($16.99) and Zapier Professional monthly ($29.99): about $127 a month in AI alone. After a month of noting what they actually used, they keep Claude Pro and move it to annual billing, switch Otter to annual, move Zapier to annual, drop Google AI Pro (Gemini came with Workspace) and cancel ChatGPT and Perplexity. New total: about $45, with nothing they used lost. The 2026 subscription price comparison is handy for this kind of audit.

Settings every solo stack needs on day one

• Training off on every consumer plan you use, checked again after major app updates.
• Two-factor sign-in on the assistant, the note-taker and the automation platform. Each one can read client material.
• A business account, not a personal one. Sign up with your business email so client work isn't mixed with your personal chat history, and so you can hand over or close the account cleanly.
• One client, one Project. Never mix two clients' files in a Project; a summary that quotes the wrong client's figures is the solo consultant's version of a data breach.
• Retention. Delete call recordings you no longer need, monthly. Keep the notes, not the audio.

A 30-day test for every tool you keep

For the first month of any new subscription, keep a one-line log each time you use it. At the end, the pattern decides for you. A filled-in illustration for a consultant trialling Perplexity Pro:

Tool: Perplexity Pro ($20/month)    Trial: 1-30 September
Uses: 7 (3 in week one, 4 across the rest of the month)
Tasks: 5 x discovery-call prep, 2 x market note sources
Would my main assistant's research mode have done it? 5 of 7: yes
Time saved vs that: maybe 15 minutes in total
Decision: cancel; revisit if market notes become a paid deliverable

A tool used three or more times a week, for work you would otherwise do slowly, stays. A tool used a handful of times, for work another tool already covers, goes, however impressive the demo was. Repeat the check every six months, because your work changes and so do the plans.

Solo consultants' questions about AI subscriptions

Is the free version of ChatGPT or Claude enough for a solo consultant?

For occasional drafting, perhaps. For daily client work, the paid plans matter less for extra features than for higher usage limits, file handling and Projects that keep each client's context together. Whatever plan you use, switch off the model-training setting before any client material goes in, because consumer plans, free and paid, rely on you doing that yourself.

Can I buy ChatGPT Business or Claude Team for just myself?

Both have a two-seat minimum, so a solo consultant would pay for a seat nobody uses: $40 a month for ChatGPT Business Standard on annual billing, for example. It can still make sense if a client contract demands business-grade data terms or admin controls. Otherwise an individual plan with training switched off is the usual choice.

Is Microsoft Copilot Pro still an option?

No. Copilot Pro stopped being sold to new customers on 1 October 2025 and support for existing subscribers ended on 1 August 2026. Its consumer replacement is Microsoft 365 Premium at $19.99 a month. For a consultancy, the business route is Microsoft 365 Business Standard plus Copilot Business, which costs less as a bundle ($23.50 a month on annual billing) than bought separately.

Should I use the API instead of a subscription to save money?

Only if you're building an automation. Subscriptions such as ChatGPT Plus and Claude Pro don't include API use, which is billed separately per token on the developer platform. For a person chatting with an assistant all day, a $20 subscription is simpler and usually cheaper; the API makes sense for a Zap or scenario processing many small jobs.

Further reads

• How Much Time Can AI Save a Solo Consultant Each Week? — How many hours this stack can realistically hand back each week.
• How Consultants Use AI to Write Proposals in Under an Hour — Put the chat assistant and note-taker to work on proposals.
• How Consultants Use AI for Client Research Before Discovery Calls — A research routine that decides whether you need a separate tool.
• Who in Your Team Actually Needs a Paid AI Licence? — The same licence question once you hire your first associate.
• Reusing Past Client Work With AI Without Leaking Client Data — Reuse past engagements inside Projects without mixing clients' data.
• How to Review an AI Tool After 90 Days: Keep, Fix or Cancel — A fuller keep, fix or cancel review for each subscription.
• Is AI Worth It for a One-Person Bookkeeping Business? — The break-even sum for a solo bookkeeper, three practice profiles with different answers, and a 30-day test before you commit to any AI tool.
• Are AI Tools Built for Coaches Worth Paying For Over ChatGPT? — What coaching platforms, AI clones and note-takers add over a $20 ChatGPT plan, a year of each costed, and the platform-closure risk to plan for.
• Is Copilot Pro Worth It for a Sole Trader? — Copilot Pro is retired. The four options that replaced it for a one-person business, priced for a year, with the traps in a single-user Microsoft account.
• Is My Business Too Small for AI? What Works for Teams of 1 to 10 — Why headcount is the wrong test for AI, what plans cost for one, three or ten people, and three small teams worked through with numbers.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: ChatGPT, Claude, Google AI, Perplexity, Microsoft 365, Google Workspace, Zapier, Make, Otter.ai and Fathom pricing pages; OpenAI's custom GPT retirement FAQ; Zoom and Google Meet help pages on meeting notes.