Michel Lieben 🧠

I'm 31. My business made > $5,500,000 in 2025.

But before that, I failed 7 businesses in a row:

Here's what I tried building...
and how each 'fail' got me closer to a 'win'


1. Started a clothing brand with my friend Thibault Selderslagh.

👎 Barely sold 15 t-shirts. Minimum order was 400. 385 tees are now collecting dust in a basement.

👍 But... I learned how to create websites.

↳ A skill I used to build the V1 of ColdIQ's website.


2. Co-Founded a raffle lottery in the UK 🇬🇧

👎 Invested 40,000€ of personal savings. Only sold £1,000 worth of tickets.

👍 But... I learned to run ads while desperately trying to sell them.

↳ We now book > 30 monthly meetings via ads and offer it as a service to clients (some closing > $2M in under a year as a result!)


3. Tried becoming an Instagram influencer.

👎 Found a hack that grew me to 29,000 followers. It then stopped working overnight. Never made a dime from IG... not even a single free product.

👍 But... I learned about social media & algorithms.

↳ Which led me to become a LinkedIn influencer instead 😆. Jokes asides, it contributed to 20,000,000+ impressions & 1,700+ inbound meetings for my company.

(And... we now offer this as a service too. Send a DM!)


4. Ran a sports betting affiliate website.

👎 Grew it to $6,000/mo. Then it dropped to 0 overnight because of regulation changes in Belgium 🇧🇪

👍 But... I learned to monetize affiliate brand deals.

↳ Today, we generate > 5-figures/mo in passive affiliate revenue.


5 & 6. Started both an MMA Betting blog AND a Crypto blog.

👎 Neither made a single $1.

👍 But... I learned SEO & Copywriting.

↳ Both these skills drove 100s of thousands of visits to ColdIQ's website.


7. Launched a Facebook Ads agency.

👎 Booked 15 outbound meetings. Closed zero clients.

👍 But... I learned cold email in the process.

↳ That skill turned into a service we delivered for > 200 B2B orgs, now generating over half a million $ per month.


Takeaway?
You win, or you learn.

Each failure becomes a lesson...
That allows you to give it another shot... better.

It's a whole process that rewards those who stick to it.

So keep pushing & good luck!


P.S: What's the worst business idea you tried?

What our $7,000,000/year funnel looks like:

(It added ~$4M ARR last year & booked > 1,500 meetings)

1️⃣ Lead Generation

We attracted ~330,000 unique visitors to our website in 2025.

They came from:

- Automated Outbound
- LinkedIn publications
- LinkedIn Ads
- Google Ads
- SEO

Software used:

FOR OUTBOUND

→ Intent Signals: PredictLeads, Common Room, Trigify.io, Clay
→ Sales Engagement: Instantly.ai, lemlist, Expandi
→ Data Enrichment: Prospeo, FullEnrich, Clay
→ Data Sourcing: Clay, Openmart, LinkedIn

FOR INBOUND:

→ SEO: AirOps, Ahrefs
→ Ads: Fibbler, LinkedIn
→ Content: Taplio, Scripe, Notion


2️⃣ Lead Capture

To convert this traffic, we rely on a few mechanisms, i.e:

- Lead Magnets & Mini-Tools, to collect email addresses.
- De-anonymization technology, to re-engage visitors.
- VSLs, to articulate our offers as clearly as possible.

Software used:

→ Visitor Identification: Instantly.ai, Midbound, Vector 👻
→ Mini-Tools: Claude Code, Lovable, Vercel
→ VSL Hosting: Wistia on Webflow


3️⃣ Lead Management

When leads don’t convert directly, we:

- Run outreach campaigns to warm leads that were scored & segmented prior.
- Nurture leads via our Newsletter & Email Marketing sequences.
- Connect on LinkedIn & nurture leads through content.

Software used:

→ Newsletter: beehiiv
→ Email Marketing: Customer.io
→ Lead Scoring: Clay, OpenAI, Claude
→ Workflow Automation: Relevance AI, n8n


4️⃣ Lead Conversion

Then all that is left is to close these leads.

This mostly happens via video conferencing on Google Meet.
Followed by proposals being sent via email.
And contracts being signed digitally.

Software leveraged:

→ Sales Enablement: Attio
→ Meeting Notes: Attention
→ Quoting/Billing: Hyperline


P.S: The strongest part of our funnel is, by far, our lead gen:

Syncing LinkedIn Ads, Content & Outbound is producing so many leads… that we don’t need to be 100% perfect lower down the funnel.

Let me know if you'd like us to run this for you!

I went viral on X sharing how our internal LinkedIn competition added $151K MRR within 90 days.

We’re running one again, but I’m changing the rules:

For context:

We gave away $5,000 (🥇), $2,500 (🥈), & $1,500 (🥉) for the winners.
And $500 to every team member who published 20+ times within the quarter.

The result: 27 new clients = +$151,000/mo.

Here are the new rules to get even better results this time:


1. We’re bumping the prizes: 🥇= $6,000 | 🥈= $3,000 | 🥉= $1,500.

I may (or may not 👀), bump these prices during the quarter + add some for 4th/5th place to keep the team motivated during the entire quarter.


2. Every participant needs to have a ColdIQ-branded banner, as well as a CTA pointing to our website

This is self-explanatory, but team impressions don’t matter if prospects can find their way to your offerings.


3. Scoring system: 1 point per like | 2 points per comment | 5 points per 1,000 impressions | 10 points per individual publication

- I want to reward “volume” by giving away points every time someone posts.
- I want to reward comments, more than likes, as they impact post reach more.
- I want points for impressions as well, since saves/shares aren’t ‘priced in’, but have a huge impact on overall reach.


4. Score multipliers: Video post = 2X on likes/comments. Original post = 1.5X on likes/comments

LinkedIn doesn’t necessarily love videos over other formats, but I’m convinced it’s the highest-leverage format out there.

Buying decisions for $25,000+ ACV products require depth… and there’s no better format than videos to achieve this.

I want the team to keep on posting carousels, infographics, text-posts & the likes of these… but want to give an extra push towards video content.

-

Also… one way to achieve performance on LinkedIn involves repurposing high-performing pieces. Most often, reposting the same piece as a colleague won’t have a negative impact on reach. So I don’t discourage the team from doing so, BUT I strongly encourage them to come up with net-new content.


5. Relative audience growth: extra rewards on posts for engagement relative to audience size.

Still trying to compute this one, but all the above scoring criteria reward larger accounts on the team.

I (70K), Alex (62K), Ivan (21K), Soheil (15K), Kenny (13K), Monika (12K) are all above 10K followers.

Hitting 100+ likes on a post is much easier than for new accounts just starting out.

Taking this into account to give a fair chance to anyone

(Also, Alex & I aren’t participating… so if you were wondering: no, I’m not planning to cash out the $6,000 prize haha!)


Loïc R. built a live leaderboard to track the competition in real-time, lmk if you’re interested to view it & I’ll share it below :)


P.S: Anyone thinking of launching a similar competition for their team this year?

What lemlist's $42,000,000 ARR funnel looks like:

(It generated +$31M in new ARR in 2025)

Their exec. team, including their CMO (Domitille de Saint-Exupéry) & their CEO (Charles Tenot) were kind enough to share it all ↓


1️⃣ Lead Generation

lemlist attracted 2M+ visits to their website in 2025.

To generate this traffic, they leveraged:

1. LinkedIn

10% of their team actively posts on LinkedIn, totalling > 300K followers.

including:
- Guillaume (46K)
- Erwan (39K)
- Charles (37K)
- Tal (35K)
- Domitille (7K)
- Yann (7K)

Software they use:

- Taplio (content)
- Figma (design)


2. Paid Ads

They invested $550K in 2025 across LinkedIn, Google & Meta Ads, producing +$4M in new ARR.

They favour Google for high-intent, bottom-of-funnel searches.

And LinkedIn for ABM campaigns targeting larger companies or to amplify their CEO's organic posts.

Software:
- Fibbler (LinkedIn ads signals)


3. SEO

- They rank #1 on Google for high-value keywords like "multi-channel prospecting tool"
- They shipped 'meta' resources that establish authority (e.g: '95 Buying Signals Directory').
- They produce multi-lingual content that ranks across several languages.
- They created multiple free mini-tools (e.g: Email Finder).

Software used:

- Semrush (Keyword research)
- Storyblok (CMS)


4. Partners, Affiliates & Influencers

They invested $60K on Influencers, both large & micro-influencers.
And allocated $450K to their partnership motion to generate +$4M in ARR.

Software leveraged:

- PartnerStack (affiliate platform)
- Crossbeam (account mapping)
- Notion (partner management)


2️⃣ Lead Capture

They converted more than 200,000+ companies into a free trial in 2025
(~12% visit-to-trial conversion rate).

To pull this off, they:

- offer a 14-day trial to their platform (no cc details required).
- let leads book demos in-app, on the website or via email.
- run reverse trials that switch users back to a $0 free plan.

Large accounts are routed to a dedicated landing page to book a demo straight away.

Software leveraged:

- Email cadence: Customer.io
- Landing pages: Webflow
- Booking: lemcal


3️⃣ Lead Conversion

Interested leads are enriched and scored based on specific thresholds:

- smaller accounts receive group demos.
- ICP fit leads receive a 1:1 demo.

They capture their sales meetings inside Claap which auto-logs deals inside HubSpot.

For closing, they leverage Katalyz, a digital sales room, for large deals with multiple stakeholders.

They generate quotes via Hyperline, which includes an e-sign process & automatically handles provisioning through their Stripe integration:

Tools leveraged:

- Deal closing: Hyperline, Katalyz
- Sales Meetings: Claap


P.S: How far is your org. from having a similar growth engine?

I spent 150+ hours trying out 62 AI Agents.

These 7 actually help you sign more clients:

1. Agent Builder: Relevance AI

↳ It helps you build custom Agentic workflows.

For example, you can build: 
→ A workflow that identifies viral posts given a list of topics/creators, & adds the ones that outperformed to a Google Sheet.

Type out inside Relevance what workflow you’d like to build 
…and it does most of the automation building on your behalf.


2. Research agent: Claygent

↳ It helps you effortlessly research prospects at scale.

Claygent is embedded inside Clay & works like an assistant you can delegate time-consuming tasks to.

For example, ask it to: 
→ Find pricing info on 1000+ SaaS companies.

Claygent will then:
- Browse the web and navigate each SaaS website.
- Navigate until it finds pricing.
- Give you the price.
- Explain where/how it found it.


3. Reply agent: Instantly.ai

↳ It monitors your inbox and drafts answers on your behalf.

Typically, with cold emails:
The faster you reply, the more meetings you book out of interested leads.

Example use-case:
→ Activate Instantly’s reply agent for your outbound campaigns.

And then, these reply agents can be trained to automatically:
- assess whether prospects were interested or not
- draft a reply, based on interest, that mimics your tone

You can run it entirely on autopilot 
…but I suggest always keeping a human in the loop for quality control.


4. Development Agent: Lovable

↳ It helps you build mini-tools without coding experience.

For example, ask it to:
→ Build an app that finds email addresses or a client reporting dashboard.

And then, after a good chunk of debugging, it creates a ‘decent’ prototype.


5. Sourcing Agent: Exa’s Websets

↳ It builds precise prospecting lists only requiring natural language prompts.

Example use-case:
→ You tell it you want to build a list of "CEOs in AI SaaS companies, from 100 to 250 employees, in NY, who raised > $1M”

The agent comes up with the filters, sources the list in real-time, provides leads that fit some of the criteria, and only charges you for leads that fit them all.


6. LinkedIn Agent: Valley

↳ It mimics human SDRs in the DMs.

Example use-case:
→ Let it run your prospecting.

Valley autonomously: 
- sources your ICP
- identifies relevant signals
- writes outreach messages
- monitors conversation & replies

... and it does that pretty impressively for a 🤖.


7. Sales agent: Attention

↳ It records your meetings & provides insights your sales team can leverage to close deals.

For example: 
→ Let Attention join your meetings.

It’ll analyse conversations in real-time, update your CRM, & suggest best next steps.

Then, it can also analyse 100s of sales conversations, and provide insights around:
“What made prospects book a meeting with your team”
“How did they hear about your company”
“What’s preventing them to buy”


P.S: Have you recently 'hired' an AI agent in your business that effectively helped you find clients?

I just crossed 60,000 followers. LinkedIn got us to $6.5M+ ARR.

Here's every tool I used to grow my following:

1. Content Ideation

→ Taplio (go.coldiq.com/taplio)

I use their in-app feed to find outlier LinkedIn posts in my niche.

You can input keywords such as: "ai sales", "outbound" or "GTM" and then filter for posts that:

- include the above 'searched' keywords
- were posted within a month ago
- had 300 likes or more

I then reverse-engineer the structure, hook & CTAs of these high-performing posts to improve my own.

→ Scripe

Also great for content ideation.

It lets you aggregate posts by author, which is another helpful way to find 'viral publications' or 'outliers' on a given topic.


2. Copywriting

→ Grammarly

As a non-native English speaker, I make lots of spelling mistakes.

I use Grammarly's extension to correct my writing in real-time.

Doesn't sound too crazy, but this could be the most helpful application in my LinkedIn tech stack.

→ Claude

I use Claude to generate hook variations, repurpose past content & brainstorm new angles.


3. Visuals & Design

→ Figma / Adobe

My LinkedIn strategy relies a lot on visuals.

When made well, they multiply the reach of your publications up to 10X.

Our designers use either Figma or Adobe's suite to design our carousels & infographics.


→ Canva

My first-ever carousels published on here were made on Canva.

It is the most beginner-friendly app to create social media visuals.


4. Video Creation

→ OpusClip

I use Opus's AI features to automatically create short-form videos from my long-form content.

The output isn't as great as that of a professional editor...
But it's cost-effective.

And multiple AI-edited videos got me 50,000+ views on LinkedIn (while only taking me 2 minutes to generate).

→ Screen Studio

Screen Studio is my favourite Mac Screen recording app.

It's great to make video walkthroughs or tutorials because it automatically zooms/focuses on the parts of the screen you're actively working on.

It makes video outputs look much more fancy than they'd be otherwise.

→ Ezgif

I use Ezgif to create .GIF files out of my short audioless video content.

If you're planning to publish a < 15-second audio-less video on LinkedIn...

I'd recommend speeding it up and converting it into a GIF file.

I'm not 100% sure why, but it outperforms in terms of reach & engagement.


5. Content Planning

→ Notion

My central content hub.
It's where I collect ideas, draft them, and prepare my posts.

Notion also serves as my content calendar.


That's the entire stack I used to go from 0 to 60,000+ followers.

It's also the tech stack I plan to get past 100,000.

What tools did I miss?

Let me know 👇

P.S: If you want to grow your account, sell a B2B product, but don't have time to post... send a DM.
We'll do it for you ✌️

Yesterday, > 1400 people watched Kenny Damian build an entire outbound campaign using Claude code.

He wrote a few prompts in the terminal, fed it a list of companies, and then...

Claude Code took over:

1. It scored and prioritised the entire company list against our ICP

2. It found verified email addresses through Apollo’s API

3. When Apollo didn’t have coverage, it automatically pulled in additional enrichment providers to fill the gaps

4. It wrote personalised copy templates on its own (based on our best-performing outreach)

5. Then it uploaded the entire campaign: leads, sequences, and copy... straight into Instantly.ai

It took fewer than 20 minutes.

This was just one of three demos we ran.

Right before that, Ivan Falco showed us how to automate ad campaign creation & 'warm' social prospecting.

Unfortunately, I can't upload the entire video here due to LinkedIn's file size limits.

If you couldn't attend:

Comment "REPLAY" to get the recording and the full playbook bundle (frameworks, SOPs, templates) so you can try this yourself.

P.S: What's a GTM play you're trying to automate using Claude Code?

I asked 62 GTM leaders running $1M+ orgs what tools they use.

One tool showed up in 71% of stacks:

MOST MENTIONED:

- Clay (71%)
- n8n (48%)
- HubSpot (39%)
- Instantly.ai(35%)
- HeyReach.io (35%)
- Apollo (35%)
- LinkedIn (32%)
- Attio (32%)
- Prospeo (29%)
- Lovable (29%)


FEATURED STACKS:

Nikola Sokolov (CEO, influencers.club) - 54 tools
↳ Clay, Instantly, HeyReach, n8n, HubSpot +49

Charles Tenot (CEO, lemlist) - 40 tools
↳ Clay, n8n, Apollo, Ahrefs, HubSpot +35

Jaspar Carmichael-Jack (CEO, Artisan) - 25 tools
↳ HubSpot, Gong, Lovable, Ahrefs, Artisan +20

Benjamin Douablin (CEO, FullEnrich) - 19 tools
↳ HubSpot, Instantly, HeyReach, n8n, FullEnrich +14

Eoin Clancy (Head of Growth, AirOps) - 17 tools
↳ AirOps, Lemlist, Clay, Common Room, Attention +12

Kai Brandt (CEO, Enginy (formerly Genesy)) - 17 tools
↳ HubSpot, Claude, Ahrefs, Lovable, Loom +12

Pierre Herubel (Co-Founder, Content Path) - 15 tools
↳ Clay, Claude, Attio, Lemlist, Webflow +10

Roq Xever (Founder, PredictLeads) - 15 tools
↳ HubSpot, Clay, Claude, Cursor, Lusha +10

+49 more stacks broken down on our "tech stacks" page.
(Link in the comments)

P.S: Any tools surprise you?

How I get > 30% replies via LinkedIn DMs

By doing the opposite of the 100+ weekly pitches I get:

CONTEXT

I sent 5M+ outbound messages for my $6.5M+ ARR Agency.

Nowadays...

Answering DMs costs me more time than running campaigns.

I'm the CEO at a 33-person company.
(I can't imagine how much outreach execs in big orgs get)

That said, I know most execs check their inbox.
And many reply.

Here's how I decide whether I'll respond or not...

And how I changed my own outreach because of this ↓


✅  DMs offering free work

Some DMs offered a free landing page rewrite or samples of already-made work.

I ran prospecting campaigns and pitched free custom lead lists.
(These were getting me more positive replies than I could handle)

My recommendation: Imagine your prospect is an existing client. What's the first thing you'll do for them? Do it already and send it over!


✅ DMs asking to be referred

Let's say you:
1. Send a message based on a current initiative of ColdIQ
2. Searched who is in charge of this initiative.
↳ I'll gladly refer you to said person

PS: You can automate this 'play' via Clay by combining intent signals & retrieving a leader/manager in a specific department.


❌ DMs asking for time on the 1st touchpoint

Meetings are my personal hell.
I'm fighting every day to eliminate as many of them as possible.

If you're "just asking 15 min of my time" to pick my brain, I won't answer.


✅ DMs from ‘known’ people

An unfair truth in cold outreach: $WHO sends a message matters even more than WHAT the message is.

If you're a well-known executive, I'll most likely reply.

I saw this firsthand as I grew my LinkedIn to 60K+ followers...

The exact same message would perform better, without changing anything.

Pro tip: If you sell B2B products, build your brand on LinkedIn.


❌ Generic templated outreach

The only way to make impersonal, generic outreach work is to send at a massive volume (millions of messages).

LinkedIn doesn't allow such volume.

So you'd be better off doing this by email.
Or better yet... not at all.


✅ DMs sharing valuable insights

Jihad explained this better than I could under one of my posts:

"
The best "free work" isn't actually work → it's insights.

I send technical audits of prospects' platforms.
→ 5-minute analysis
→ 3 specific optimization opportunities
→ No strings attached

Example: "Your checkout flow has a 23% drop-off at payment verification. Here's why + 2 quick fixes."

- Takes me 10 minutes.
- Shows immediate value.
- Demonstrates expertise without asking for anything.

The key: Make it so valuable they'd typically pay for it.

Generic "free consultation" = everyone does this.
Specific actionable insight = rare.
"

Solving prospects' issues before they pay you a dime is the fastest way to gain their trust.


🔘 Videos or Voice Notes

I listen to voice notes or watch videos.

It annoys people.
But if they're well done, they'll grab attention.

What's a DM you receive you could NOT ignore?

Share it 👇

Posting on one LinkedIn account brought us new clients.

24 accounts added $153,000 MRR in 90 days:

We collectively hit 'publish' 581 times...
which added up to:

- 43,473 reactions
- 28,130 comments
- +34,023 new followers

But most importantly:

+ 27 new clients
+ $153,000 in MRR

… in the span of 87 days.

And our sales kept hearing the same thing from prospects:
“We see you guys everywhere”

Here’s how we did it 👇


1️⃣ We ran an internal LinkedIn competition

Your team may want to post on LinkedIn.
But if it's not in their job responsibilities… good luck.

We ran a competition & gave away the following prizes:

🥇 $5,000
🥈 $2,500
🥉 $1,500

… & $500 for every person posting > 20 times during the quarter.


2️⃣ We removed as much friction as possible around 'posting'

Successful posts have a “recipe”.

It's usually a combination of:

→ A proven angle or trendy topic
→ Breathable, structured writing
→ A strong hook
→ A nice visual

But:

Identifying proven topics isn’t easy.
→ We put Taplio & Scripe at their disposal to help the team see what works.

Brainstorming hooks can be hit-or-miss.
→ I tell my team to reuse variations of mines that outperformed.

Designing visuals takes time.
→ We hired designers to handle this part, so that our team can focus on the ‘writing’ as opposed to the ‘packaging’.


3️⃣ We heavily invested resources to train the team

Making it work required 1:1 sessions in which we:

- Built a posting calendar
- Rewrote posts & hooks.
- Brainstormed ideas

And group sessions during which we:

- Shared best practices.
- Brought LinkedIn experts (e.g, Pierre Herubel & Lara Acosta) to teach our team.


4️⃣ We hired people motivated to grow their brand

We used to ask applicants whether they wanted to grow on socials.

At the same time, many new joiners heard about us from a LinkedIn post.

If WE attracted them, they were typically willing to 'publish'.

For the story:
Kenny Damian learned about ColdIQ from a post by Monika Grycz 💌.

He was already posting by day 2 of joining.

He had fewer than 1,000 followers in 2025 a year ago.
He’s recently reached 10,000.


5️⃣ We led by example

We granted $500 prizes to team members who published 20+ times.

Meanwhile, I published > 60 times in the same timeframe.

If you’re going to ask your team to post
You should demand it from yourself as well.

Being near ~70,000 followers makes me more credible when I tell my team that I’m sure they can reach 10,000.

Plus, we also have 13 team members > 5,000 followers:

- Michel: 67,752
- Alex: 60,601
- Ivan: 21,090
- Yevhen: 16,774
- Soheil: 14,217
- Monika: 11,300
- Kenny: 10,633
- Mai-Lan-Lan: 7,970
- Jan: 7,089
- Dujam: 5,887
- Juan: 5,823
- Harry: 5,381
- Louis: 5,258

Every successful audience growth builds conviction for the rest of the team.

P.S:

If you'd like to get your team 'successfully' posting in 2026...

I just recorded a video that details how we made it work.

Let me know if you'd like to view it & I'll send it over :)

These 4 GTM agents will save you 5 hours every day:

They automate:

→ Content creation
→ Market research
→ Prospecting

Here’s how they work:

AGENT 1: AUTOMATE LEAD RESEARCH

This agent captures company profiles, enriches them & scores them on autopilot.

Step-by-step:

- GhostGenius sources company profiles
- Agent monitors credibility signals (follower count, high activity, active website) 
- GPT scores each profile against our ICP criteria.
- Contacts are deduped so they’re never reached out to twice.

Qualified validated prospects are sent directly to a Google Sheets for prospecting.


AGENT 2: LOCAL OUTBOUND ENGINE

Apify extracts information about local businesses from Google Maps.
But this raw data is typically unstructured, messy & outdated.

Thus, through the Agent:

- Google Search API fills incomplete metadata.
- Each website is scraped for keywords, tech stacks, & offerings.
- GPT enriches every business profile & classifies these.
- Additional info (i.e: emails, phone numbers, contact links) is extracted.

The data is then exported to Airtable or Google Sheets.
Initial low-quality data turns into sales-ready records.


AGENT 3: CONTENT GENERATOR

Drop a content brief into a Google Form.

Get auto-generated:

- Copy for your LinkedIn carousels, slide-by-slide.
- Visual briefs for your designer.
- LinkedIn publication drafts.
- LinkedIn (or social) Hooks
- CTAs

Drafts are automatically added to your Notion content calendar.

They need a 10 to 15-minute manual edit, and they’re good to go.


AGENT 4: REDDIT INTEL SOURCING

Many GTM teams ignore forums & communities.
But they’re typically goldmines for learning about your prospects' pain points.

Reddit has communities in almost every niche.
But it’d take lots of time to read through all these.

Thus, the agent, based on a website/domain:

- Identifies ICP signals.
- Extract relevant threads via Reddit API
- Summarizes pain points, language used by prospects, frustrations & what they seem to care about most.
- Sends a daily report to your inbox.


THE RESULT: MORE PIPELINE

These steps save a lot of manual time, which you can now allocate to going in-depth into more worthwhile, high-value activities.

All these automations are built inside n8n.

Which one is most valuable to you?

AI agents are replacing human employees.

Funny enough, they operate in very similar ways:

The smartest agents don’t work alone.
They work in teams.

Some behave like assistants.
Others as managers.

And as with any sort of team, there are different ways to structure their collaboration.

To illustrate this:
We can look at how 3 multi-agent setups would handle the same task…

For example: Scheduling a quarter business review with an important client.


1️⃣ Hierarchical setup

A manager agent assigning tasks to its agent team members.

1. The manager receives a task = Schedule a business review.

2. It tells a subordinate agent to check potential times to meet with the clients.

3. Another agent checks product usage from the client. "Has usage recently dropped? Is engagement still strong?"

4. A 3rd agent drafts an email based on client availability & product usage:

"Hey Achilles, it's been a while - wanted to ensure you're using the product to the fullest. How about a quick chat next week, Wednesday?”

5. The manager reviews everything and sends the email.


2️⃣ 'Human-in-the-Loop'

A smart assistant tapping your shoulder when needed.

1. The agent drafts the business review email and finds some available times.

2. But… it sees the client gave negative feedback 2 weeks ago.

3. Before sending the email, it asks you: “Do you want to review the email before I send it?"

4. You make small tweaks, and you send it.


3️⃣ 'Sequential' setup

Like a factory or an assembly line where each person does one task and hands it off to the next.

1. 1st Agent asks: "Is a quarterly business review due?" based on the client's last meeting time.

2. 2nd Agent pulls client's usage data.

3. 3rd Agent drafts a short email.

4. 4th Agent sends the email to book the meeting.


So, when should you use which?

As a rule of thumb:

- Hierarchical setups work well for complex, multi-step tasks.

- Human-in-the-loop setups are recommended for high-stakes, sensitive tasks.

- Sequential setups for ordered workflows with clear dependencies.


P.S: What's one process in your team that would run better...
with some smart agents working behind the scenes? 👇