This review is part of a larger series of LinkedIn newsletters titled The Human Side of AI: Cutting through the AI noise to show you how AI can be a powerful tool for your creativity, efficiency, and strategy.
The most useful GenAI tools for research operations are not always the best talkers. They are the ones that connect to the rest of the stack, and increasingly, the ones that act on it. Two shapes have emerged. Connectors like Lindy and Make move data between tools and add intelligence at the seams. Agents like Manus take a goal and run it end to end. Here is where each fits for a B2B research team.
What’s new: When we first reviewed this in July 2025, Lindy was the example and Make was a promise. This update delivers the Make section that was coming, adds Manus, an autonomous agent that does the work rather than only connecting tools, and refreshes Lindy. The biggest change on Lindy is pricing: the original left Cost as “TBD,” and it is now a usage-metered model worth understanding before anyone commits. Lindy’s data-privacy stance, by contrast, has largely held up.
Lindy
Lindy links existing tools, spreadsheets, documents, Gmail, and CRM systems into end-to-end workflows without code, and routes the thinking through models like Claude, ChatGPT, or Perplexity. The research use cases from the original review still hold. If a team drops interview transcripts from a tool like Fathom into a shared sheet, Lindy can watch that file, summarize each new entry, and route it to the right stakeholder. Point it at a list of companies or products and it can run a secondary research pass, then draft the report or email from what it finds, triggered whenever a new row lands. It also works for trend monitoring, a more flexible Google Alerts, and for meeting prep, pulling past threads, LinkedIn profiles, and CRM notes into one brief before a client call.
Since 2025, Lindy has leaned harder into the personal-assistant role: an iMessage and SMS front end, voice agents, browser-driving computer use, and a catalog now in the thousands of integrations. The research-workflow value is still there; there is simply more product around it.
On data handling, the reassurance the original review pointed to has held up, which is not something every tool in this series can say. Lindy states that customer data is never sold, shared, or used to train models, and it backs that with enterprise security positioning that includes SOC 2 Type II, GDPR, PIPEDA, and Enterprise HIPAA support with a signed BAA. For client-confidential transcripts or product roadmaps, that posture matters.
Cost is where the picture changed. The original left it as “TBD,” and it is knowable now. Lindy runs on usage-based pricing, with Plus at $49.99 a month, Pro at $99.99, and Max at $199.99, plus custom Enterprise pricing. Current plans are billed monthly and paired with a 7-day free trial rather than a permanent free tier. Every action draws from a usage pool, and the draw scales with task complexity, so a research-heavy workflow can burn through an allocation faster than the sticker price suggests. Recent pricing guides also note overages that can run at about double the baseline credit cost. The practical move is to run real workflows through the trial first and turn on usage warnings before building anything mission-critical.
Ratings
| Dimension | Rating | Rationale |
|---|---|---|
| Usability | 4.5 / 5 | Smooth interface with thoughtful defaults and quick integration setup; natural-language agent building is genuinely fast. |
| Power | 4.5 / 5 | Automating interview summarization, trend monitoring, and meeting prep saves research teams hours a week. |
| Flexibility | 4 / 5 | Not a fully open-ended agent, but plenty of room to customize within specific tasks and flows, now across thousands of integrations. |
| Cost | 3 / 5 | No longer “TBD.” Credit-based metering makes budgeting harder: complex research tasks can burn credits fast, and overages may run about 2x. Model real usage in the trial first. |
Make
Most tools in this series help teams do the research. Make does something different. It is the connective tissue between everything else. Make is a visual, no-code automation platform that links more than 3,000 apps, and increasingly AI models, so data moves between tools without anyone copying and pasting. For a research team buried in the unglamorous work of getting data from one system to another, that earns a real look.
A visual builder that does not require a developer. Make’s drag-and-drop canvas lets users build a workflow, trigger to transform to destination, without writing code. Set up a scenario once, like pulling new survey responses from a form into a shared repository, and it runs on its own from then on.
AI as a first-class citizen, not a bolt-on. Make has moved past being a plain integration tool. AI Web Search brings live data into a workflow, and Make AI Agents orchestrate agentic workflows across connected apps. For a research team, that means a pipeline where a model reads incoming data, makes a routing decision, and acts on it without anyone shuttling files between platforms.
It connects the rest of the stack. Slack, Airtable, Google Sheets, HubSpot, NetSuite, and Notion all show up as native apps, alongside thousands of others. If a client wants competitive intelligence pushed to a Slack channel the moment it is flagged, or survey data auto-formatted into a shared drive, this is what Make is built for.
Where it could improve. Pricing runs on a credit system, and every action a scenario performs consumes credits, so a team running dozens of scenarios across engagements needs real monitoring discipline to avoid a surprise bill. The AI features are more mature than a typical beta label would suggest, but they still deserve deliberate testing before anything client-facing rides on them. And the visual builder is approachable for simple automations while routers, iterators, and error handling add a genuine learning curve once messy real-world data enters the picture. Company single sign-on is also tied to higher-priced plans, which matters under strict client security requirements.
Ratings
| Dimension | Rating | Rationale |
|---|---|---|
| Usability | 4 / 5 | The canvas is intuitive for straightforward automations; multi-branch workflows take longer to master. |
| Power | 4.5 / 5 | More than 3,000 app integrations plus AI workflow and agent orchestration let Make sit at the center of a research team’s operating stack. |
| Flexibility | 4.5 / 5 | Works across research operations, marketing, sales, and internal ops without forcing a single workflow pattern. |
| Cost | 3.5 / 5 | The free tier and low entry point are appealing, but credit consumption on complex scenarios can get expensive and unpredictable at scale. |
Make will not analyze data or write a report on its own. It removes the repetitive plumbing that slows research operations: the manual movement of files, alerts, and structured data between tools. The AI can move the data and flag what needs attention. Deciding what the finding means still takes a researcher.
Manus
Most tools in this series are assistants. You ask, they answer, and turning that answer into something usable is still the user’s job. Manus works differently. Give it a goal, and it plans the steps, opens a browser, runs code, and hands back a finished file. It runs inside a sandboxed virtual machine with real browser and file-system access plus a terminal, a step past chatbot plugins that only simulate tool use. Give it an open-ended brief, like “compare the top 10 CRM tools for mid-market B2B on pricing and features, deliver as Excel,” and it can open tabs, pull data from vendor sites, and return a structured spreadsheet ready for review.
Key strengths
- Autonomous multi-step research. Point Manus at a research question and it plans, browses, and synthesizes across sources without a prompt at every step. That is useful for the sprawling competitive or vendor-landscape work that normally consumes an analyst’s afternoon.
- Deliverables, not just answers. Output can arrive as a finished Word doc, Excel file, slide deck, or live webpage, so the step of turning an answer into something client-ready mostly disappears.
- Wide Research mode. A parallelized mode built for large-scale scans, useful when a study calls for surveying dozens of sources rather than three or four.
- Take Over mode. When Manus hits a CAPTCHA or MFA prompt mid-task, it pauses and hands control back instead of failing silently, then resumes once the block clears.
- Manus Desktop and My Computer. Local-access modes let the agent read and edit files and run installed tools on a user’s machine, typically behind an approval step, for researchers who want it working against local archives rather than only the cloud sandbox.
Where it could improve
Pricing is the biggest friction point, and the free tier is thinner than it looks. Free usage is constrained, paid Pro and Team tiers remain credit-based, and a single complex research task can eat a substantial chunk of a monthly allocation. Teams need guardrails to keep costs predictable, and anyone publishing exact Manus plan limits should re-check the live pricing page before relying on them.
Reliability on longer, messier tasks is improving but not yet set-and-forget. Plan to review its work on anything client-facing and refine prompts over time. Most execution also still happens in Manus’s cloud sandbox, so organizations handling sensitive or regulated data should review its data-handling and retention practices before relying on Desktop or Team features.
The ownership picture needs a clear-eyed look. In April 2026, China’s National Development and Reform Commission ordered Meta’s roughly $2 billion acquisition of Manus unwound. Bloomberg reported that by June, Meta had completed an operational split and cut off data sharing between the two companies, and that Manus’s founders were exploring roughly $1 billion to buy it back. Manus continues operating as a standalone service, but the acquisition is better understood as unwound rather than pending, and where the company lands next is still an open question worth tracking before anyone builds it into a core workflow.
Ratings
| Dimension | Rating | Rationale |
|---|---|---|
| Usability | 3.5 / 5 | The agent model is powerful but different from a chat interface. Monitoring credit burn, reviewing multi-step plans, and clearing verification gates adds friction a straightforward assistant does not have. |
| Power | 4.5 / 5 | Genuine autonomous execution across research and file generation, with real browsing built in, plus finished-deliverable output that saves research teams real time. |
| Flexibility | 4 / 5 | Handles open-ended web research, spreadsheets, decks, and local file access well, but is still better suited to discrete, well-scoped tasks than to replacing a full research stack. |
| Cost | 3 / 5 | Credit-based pricing and variable task complexity make budgeting harder than with flat-fee tools. Serious use tends to push quickly toward a paid tier. |
Manus is one of the clearer signs that AI is moving from answering questions to doing the work. For the specific pain of secondary and competitive research, that shift is worth testing now, with a human reviewing the output and the corporate picture in view.
In sum
Research operations is changing, and the workflow-driven tools are why. Connectors like Lindy and Make pull the plumbing out of the day: the file movement and the status alerts that used to eat someone’s afternoon. Agents like Manus go a step further and take on the doing, returning a finished deliverable instead of a prompt to act on. None of that is standing still, either. Lindy’s pricing model shifted, Make keeps promoting its AI features toward core, and Manus is working through an unwound acquisition, so any single verdict here is good for this quarter, not forever.
What does not move is where the judgment sits. These tools can move the data and run the scan, and the newest can draft the first pass. Deciding what a finding means, and what to do about it, still belongs to a researcher. The future here is not human or AI. It is Human + AI, with the plumbing and the grunt work handed off so the thinking has more room.
At Cascade Insights®, we’re testing these workflow tools in real B2B research operations as they mature; if you’re deciding where they fit in your own stack, that’s a conversation we’re glad to have.
Last updated: 7/14/2026