9 Claude.ai Features That Replaced 4 SaaS Tools (No Code)

By Lazar Milicevic · Published June 9, 2026 · 8 min read

Most Claude Pro users treat the $20 seat like a chat window. They miss that the same browser tab — no API, no CLI, no developer setup — replaces a writing tool, a knowledge base, a transcription service, and a part-time VA. Here's the exact configuration I run with solopreneur clients on day one.

Projects: stop pasting context into every chat

A Project is a container that holds custom instructions plus uploaded reference files. Every chat inside that Project inherits both. This is the single biggest leverage point in the entire product and almost nobody configures it past the default.

For a small invoicing business I work with, we built a Project called Customer Support. Inside it:

• Refund policy (PDF)
• Pricing sheet (PDF)
• Top 40 support questions with approved answers (Markdown)
• Instructions block

The instructions block looks like this:

You are the support agent for [Company].
- Reply in the customer's language (detect from their message).
- Never promise a refund outside the policy in refund_policy.pdf.
- Quote prices only from pricing.pdf.
- Sign off with "Marko" — never "Claude" or "AI assistant".
- If the answer is not in the uploaded docs, say so and offer to escalate.

Drafting a reply went from ~12 minutes of context-switching (open policy, open pricing, write reply) to ~2 minutes (paste customer email, copy reply). One Project per business function: Support, Sales, Content, Bookkeeping. Don't dump everything into one.

Account-level custom instructions: the 20-second tax

Project instructions tell Claude how to behave for a specific job. Account-level instructions tell it how you personally like to receive information across every chat — including one-off conversations outside any Project.

Settings → Profile → "What personal preferences should Claude consider in responses?"

Mine:

- Skip pleasantries. No "Great question!" openers.
- Answer first, reasoning second.
- Plain text. No bullet lists unless I ask.
- Assume I'm a senior engineer. Don't explain basic concepts.
- If I'm wrong, say so directly.

Set once. Save ~20 seconds on every conversation forever. If you average 30 chats a day, that's 10 minutes a day back, or roughly 60 hours a year.

Styles: kill the $49/month writing tool

Styles live in the bottom of the chat box and most users never open the dropdown. A Style is a saved voice-and-formatting preset you click on per conversation.

I run three:

• Client Proposal — formal, structured, no contractions, executive-summary first
• Newsletter — short sentences, conversational, one idea per paragraph
• Technical Docs — imperative voice, code blocks, no marketing language

To create one: click the Styles dropdown → Create Style → paste 3-4 samples of writing in that voice. Claude reverse-engineers the tone. You don't need to describe it in English.

This single feature let one client cancel a $49/month AI writing subscription. The Style picker is literally a one-click replacement for prompts that say "write this in a friendly but professional tone with short sentences."

Artifacts: client deliverables without the Google Doc ping-pong

Whenever Claude generates code, a document, a table, or an HTML preview, it pops out into a side panel. You iterate in the chat, the Artifact updates in place. The chat history stays clean.

The use case nobody talks about: client deliverables as live HTML.

Generate a one-page proposal for [Client] as an HTML artifact.
Sections: problem, proposed solution, deliverables, timeline, price.
Style: clean, single column, max-width 720px, system fonts, no JS.
Use the brand color #0A66C2 for headings only.

The client gets a live preview in the same chat session. I tweak in plain English ("make the timeline a table, drop the price by 10%"), the Artifact re-renders. Export as HTML, send. No Google Doc with 14 comment threads.

Same workflow for SOPs, onboarding checklists, one-page reports.

File analysis: the part of Claude that replaces a VA

This is where you stop paying someone $300/month to do data entry.

Drag a CSV of 200 invoice line items into the chat:

Categorize each row by expense type (software, travel, contractor, 
office, marketing, other). Flag duplicates where vendor + amount + 
date are within 2 days of each other. Return a clean CSV with a 
new "category" column and a "duplicate_of" column.

Runs in ~30 seconds. Accuracy on structured data sits in the high 90s. For PDF invoices, the same pattern — extract vendor, date, total, line items, dump to a table. Batch 10-20 PDFs per chat.

One client was paying a VA $300/month for exactly this. We kept the VA for one hour a week to spot-check Claude's output and reassigned the other 19 hours to actual customer work.

Where file analysis breaks down

• Scanned PDFs with bad OCR — feed it through a real OCR step first
• CSVs over ~5MB — split into chunks
• Highly unstructured handwritten forms — still needs a human

MCP connectors: Gmail and Drive without Zapier

This is the newest piece and the most underused. Settings → Connectors → toggle on Gmail, Google Drive, Google Calendar. No API keys, no OAuth nightmare, no Zapier middleman.

Now this prompt works inside a normal chat:

Read my last 20 emails. Summarize what needs a reply today.
For each one, draft a response in my "Casual Newsletter" Style.
Skip newsletters and shipping notifications.

For a consulting client this replaced ~90 minutes a day of inbox triage. He still sends the replies manually (intentional — he wants final say), but the drafting is gone.

Pair it with the Drive connector and you can pull a contract template from a folder, fill in the client name and terms from an email thread, and get a finished draft in one chat. That's three tools collapsed into one tab.

Web search + Computer Use: research and browser actions

Web search is a per-conversation toggle. Combined with a Project that already knows your business, you get research filtered through your brand context:

Find three competitors who launched a feature similar to our 
[X] in the last 60 days. Summarize their positioning, pricing, 
and how it differs from ours (see brand-brief.pdf in this Project).

That's a $30-50/month research tool replaced.

Computer Use (agent mode in some accounts) is still rolling out and not as reliable as the rest of this stack. When it works, Claude drives the browser — clicks, types, fills forms. I use it for lead research: open a list of 20 company URLs, pull the founder name and a relevant detail from each About page, dump to a table. Treat it as "useful but supervise it" — not "set and forget."

The actual replacement math

Real numbers from one client's stack before and after:

• Jasper / writing tool: $49/mo → replaced by Styles
• Notion AI / knowledge base lookups: $20/mo → replaced by Projects
• Otter.ai transcription: $17/mo → replaced by file upload + summarize
• Part-time VA for data entry: $290/mo → replaced by file analysis
• Total killed: $376/mo. Kept: one $20 Claude Pro seat.

The catch: this only works if you set it up like a system, not like a chat window. Spend two hours on day one building 3-4 Projects, 2-3 Styles, account-level instructions, and toggling the right connectors. That two hours pays back in the first week.

Why bizflowai.io helps with this

A lot of what I build for clients at bizflowai.io is exactly this layer — taking the Claude Pro seat they already pay for and wiring it into their actual operations. Project setup with their real documents, Styles trained on their existing writing, connector configuration for Gmail and Drive, and the SOPs that tell their team which Project to open for which task. No new subscriptions, no custom code for the parts that don't need it. Custom automation only comes in when the browser stack genuinely can't do the job.

Frequently asked questions

What is a Claude Project and why does it matter?

A Claude Project is a container inside the Claude web app that holds custom instructions plus uploaded reference files, and every chat started inside that Project automatically inherits both. Instead of pasting your brand voice or FAQ document into every conversation, you configure it once. For example, a Customer Support Project loaded with refund policy, pricing, and approved answers can cut reply drafting from twelve minutes to two.

How do I create a custom Style in Claude?

Click the Styles dropdown at the bottom of the chat box, hit Create Style, then paste in three or four sample pieces of writing in the voice you want. Claude reverse-engineers the tone and saves it as a one-click preset. You can build multiple Styles — for example, a formal proposal Style, a casual newsletter Style, and a technical documentation Style — and switch between them without rewriting prompts.

What's the difference between account-level instructions and Project instructions in Claude?

Account-level custom instructions are global and apply to every chat across your Claude account, telling Claude how you personally prefer to receive information. Project instructions only apply to chats inside a specific Project and define context for that workflow. For example, your account instructions might say skip pleasantries and answer first, while a Project's instructions define a support agent persona with policy rules.

How can Claude replace a virtual assistant doing data entry?

Claude's file analysis feature lets you drag a CSV or PDF directly into the chat and ask it to categorize rows, flag duplicates, or extract fields like vendor, date, and amount into a clean table. It processes structured data in about thirty seconds with high-nineties accuracy. One client replaced a three-hundred-dollar-per-month VA who was doing exactly this invoice categorization task.

What are Artifacts in Claude used for?

Artifacts are a side panel where Claude outputs code, documents, tables, or HTML previews so you can edit and iterate without cluttering the chat history. Beyond coding, they're useful for client deliverables — for example, generating a one-page proposal as an HTML Artifact, showing the client a live preview, tweaking it in plain English, and exporting it without Google Doc back-and-forth.

Want more like this?

I publish practical AI automation, GenAI engineering, and faceless content workflows on YouTube every week.

Subscribe to bizflowai.io on YouTube — never miss a new tutorial.

Planning an AI automation project or need a second opinion on your architecture?

Connect with me on LinkedIn — Lazar Milicevic, GenAI Engineer & bizflowai.io Founder.

Visit bizflowai.io for our services, case studies, and AI consulting.