Fleets by Mridu AI

The AI SDR that reads you before it writes to you.

Visit Website
August 27, 2026 From 0 connections to 35% reply rate: How I built Fleets

The problem: I was the worst salesperson at my own company.

I'm Adarsh, founder of SolveEase. For the first year, I did all our B2B outreach myself. It looked like this:

- 3 hours finding 20 people on LinkedIn

- 2 hours "personalizing" messages that all sounded identical

- Responding to replies 3 days late because I was building the actual product

- Losing leads because I forgot to follow up

I was the bottleneck. And I couldn't afford a good SDR.

What I built:

Fleets — an autonomous LinkedIn outreach agent. But not the "mail merge with AI" kind you've seen.

Here's what actually happens when you run a campaign:

1. You describe your ICP in plain words

2. The agent searches LinkedIn like a human would — scrolling profiles, reading posts, understanding context

3. Trained models qualify every lead before outreach. Not keyword matching. Actual fit scoring.

4. It writes one message per person — referencing their recent posts, their company, their role. No templates.

5. When someone replies, the agent handles the first exchange — answers basic questions, qualifies interest, then hands the warm conversation to you

The stack:

- Playwright for browser automation (no API, real browser sessions)

- OpenAI for the reasoning layer

- Supabase for data + realtime dashboard

The numbers from a dead profile:

We started a brand new LinkedIn profile. Zero connections. Warmed it for 15 days. Then ran Fleets.

| Metric | Number |

| Profiles scanned | 1,818 |

| Connection requests | 160 |

| Accepted | 74 (46%) |

| Leads contacted | 59 |

| Replies | 17 (29%) |

50%+ acceptance rate. 35%+ reply rate. From nothing.

The thing I'm most proud of: Managed profiles.

I didn't want to risk my own LinkedIn. So we built a system to warm up fresh profiles and rent them to users. You get a dedicated, pre-warmed account. Your personal profile is never touched. Same agent, same results, zero exposure.

Pricing (billed in INR, serve globally):

- Starter: ₹4,999 (~$60/mo) — 1 profile

- Growth: ₹12,999 (~$155/mo) — 3 profiles + dedicated success manager

- Pro: ₹21,999 (~$265/mo) — 6 profiles

7-day free trial. No card required.

What I learned:

- Quality beats quantity by miles. Our reply rates are high because we say no to 90% of profiles.

- Playwright > API for anything social. APIs get you banned. Browsers keep you alive.

- Founders shouldn't be doing first-touch outreach. It's a waste of founder time.

What I need help with:

We're India-based and most of our clients are Indian. I want to expand to US/EU markets but I'm not sure how to position this for those audiences. If you've sold B2B SaaS globally, I'd love your thoughts.

Also — if you've tried LinkedIn automation before and hated it, tell me why. I want to make sure we don't become the thing we set out to replace.

https://la.mridu-ai.com/

10 Comments

  1. 3

    one thing that trips up a lot of India-based saas going global is the pricing page itself. billing in inr with a usd estimate sounds fine internally but us/eu buyers want to see a clean number in their own currency without doing conversion math in their head. that hesitation isn't about the price, it's about not trusting a number that isn't "theirs" yet. worth testing localized pricing before anything else on the expansion list

    1. 1

      this is exactly the blind spot i needed called out — thank you.

      you're absolutely right. it's not the price, it's the micro-friction of "let me check this conversion" that breaks trust in the moment. we had a us prospect drop off at checkout last week and i couldn't figure out why. now i know.

      we're a/b testing stripe's localized pricing this week — clean usd/eur numbers upfront, still billed through our indian entity. will report back on conversion difference.

      really appreciate you taking the time to write this out.

      1. 1

        that's a smart move testing localized pricing first. we ran into the same wall building worldesk, it's a free fx rate verification tool, no sign-up needed, and it turned out the conversion confidence mattered way more to buyers than the actual rate itself. happy to have our fx specialist take a look at what you're seeing with the stripe test if that'd help

  2. 2

    You could comment:

    Strong numbers, especially the qualification-first approach. Saying no to 90% of profiles is probably doing more for reply rate than just automating more outreach.

    For US/EU, I’d position Fleets less as “LinkedIn automation” and more as an AI SDR that researches and qualifies before reaching out. One concern I’d have is long-term account/platform risk with browser automation and managed profiles, so trust and transparency around that will matter a lot.

    1. 1

      haha thanks man — means a lot.
      "LinkedIn automation" sounds like a spam tool. "AI SDR that researches first" is actually better. We're rewriting the homepage hero with that framing this week.

      On the risk side — totally fair concern.

      Here's exactly how we handle it:
      - Every user gets a dashboard kill switch (one click, immediate pause)
      - Managed profiles are legally separate with usage agreements you can audit
      - We publish our safety thresholds publicly (daily limits, IP rotation, etc.)

  3. 1

    35% reply rate is impressive. I’m curious how much of that comes from personalization versus the targeting itself. Did you find a point where more personalization stopped improving replies?

  4. 1

    35% reply rate from zero connections is impressive.

    What do you think made the biggest difference — the targeting, the message itself, or the way you followed up?

    I’m currently learning a lot about early-stage outreach, so I’m especially interested in what actually moved the needle rather than just increasing volume.

  5. 1

    schmackin dude..))

  6. 1

    Last time I used a Linkedin Automation and my account got banned. These things never work.

    1. 1

      Fleets stays inside LinkedIn's published weekly limits for connections and messages. It paces activity across the day (no burst spam), runs from a dedicated IP with human-like timing, and you can pause anything instantly from the dashboard.

      We also offer managed profiles — dedicated, pre-warmed accounts that are completely separate from yours. Most of our clients start there.

      If something ever looks off, one click pauses everything.

About

Most LinkedIn tools automate sending. Fleets automates thinking first. Our trained models qualify every lead before a single message is written—no templates, no merge tags. Built as our own in-house B2B sales system.