I’m building a 24/7 AI agent that runs workflows for founders — looking for feedback

Sounds promising. I wish you both good luck!

Love it, but “ automated follow-ups” makes me nervous

if i'm handing an agent access to my social profiles, it had better run on my machine....

Interesting idea.
How does the AI decide which tools or services to use when building a workflow?

Hi! Great idea! My question is: can you scrape LinkedIn for Leeds? If not, then how and where do you search for Leeds? Creating a personalized email should not me so hard, but lead digging - I wonder how do you approach that?

Interesting concept. The 'describe it and it runs' pitch is compelling but the hardest part is reliability, workflows that fail silently are worse than manual. How are you handling error recovery and notifications when something breaks?

Love the workflow automation angle, especially the “describe the task → agent runs it” model.

Curious where you're seeing the most traction so far, outreach automation or internal workflows?

I’ve noticed founders tend to adopt tools faster when the first workflow solves a painful daily task.

cool stuff

I like this idea a lot.
“Describe the task and the agent builds the workflow” is a powerful concept.
For founders constantly testing and learning new tools, it could save a lot of wasted time. Even simple automation tools have a learning curve.
I recently set up automated emails with Brevo. They try to make it simple, but even then it takes time to figure things out.
Great concept. Keep plugging away.

This is interesting.
Many founders struggle with tools like Zapier because workflows become complex quickly. If your agent can translate natural language into reliable workflows, that could remove a lot of friction.

This is really interesting! The 'describe the task → AI builds it' approach removes a huge barrier for non-technical founders. How does Sendlume handle cases where the AI misunderstands the workflow? Is there a way to review or edit before it runs?

Interesting idea. The “describe the task → AI builds the workflow” approach feels like a natural evolution from traditional automation tools where users still need to manually design everything.

One workflow I’d personally love to automate is early-stage user research: collecting feedback, summarizing responses, and identifying recurring pain points from conversations with users.

Curious about one thing: in your early tests with founders, do people usually come with a very clear workflow in mind, or do they discover possible use cases only after experimenting with the agent?

Also nice progress getting the MVP to ~90% and already building a waitlist.

Nice progress on the MVP. The idea of describing a task and letting the AI build the workflow is very compelling.

I think document automation could be a strong use case — especially workflows around invoices, statements, and other structured documents.

Founders orchestrating AI agents 24/7 is genuinely the next wave — love that you’re building in this space.

One thing I’d push on from feedback experience: the prompt that defines what the agent does is usually the fragile part. Agents fail not because the orchestration is bad, but because the instructions are ambiguous or structurally inconsistent.

I built flompt to tackle exactly this — a visual prompt builder that decomposes prompts into 12 semantic blocks (role, objective, constraints, examples, output format, etc.) and compiles them to Claude-optimized XML. Useful for getting agent instructions tight before you scale them.

A ⭐ on github.com/Nyrok/flompt would mean a lot — solo open-source founder here 🙏

Interesting concept I love this especially the idea of building workflows just through conversation instead of traditional automation builders.

One workflow I’d personally want an AI agent to handle is podcast-to-content repurposing for SaaS companies.

For example:
• Monitor when a new podcast episode drops
• Transcribe and extract key insights
• Turn them into LinkedIn posts, X threads, and short-form content
• Identify quotable moments for thought leadership
• Draft newsletter snippets or blog outlines
• Schedule or prepare content for distribution

This kind of workflow currently requires multiple tools and several hours of manual work.

If an AI agent could handle that end-to-end, it would be extremely valuable for SaaS founders trying to build distribution from podcast content.

Curious are you focusing more on internal automation, or also creator/marketing workflows like this?

Autonomous workflow platforms solve a real pain — the 20–30% of founder time that goes into tasks structured enough to automate but complex enough that basic Zapier flows break.

'Automated follow-ups' is one of the highest-value starting points. The harder problem is the stop condition: when should the agent not send the follow-up? For payment recovery sequences specifically, sending an email to a customer who already paid is worse than not sending it at all.

That's the edge case we had to solve at tryrecoverkit.com — the D+1/D+3/D+7 sequence for failed Stripe payments only delivers value if it correctly halts the moment the payment succeeds. Curious how Sendlume handles state management across multi-step sequences where an external event (payment success, customer reply, cancellation) should change the desired next action mid-sequence.

Nice project.

Do you see this evolving more as an AI workflow builder or as a fully autonomous growth agent over time?

Curious about the long-term vision.

If it reliably handles repetitive tasks, that could be very valuable for founders.