Anas Duksi

Electrical Engineering Intern

As an electronics engineer, I got hooked on systems, logic circuits, models, things that click and scale. I was fascinated by how systems work when they do! 
As a visionary, being able to imagine and understand systems, that obsession stuck..

Now I build AI agents and automations that plug into GTM and Revops

Been deep in SaaS GTM for years - outbound, enterprise deals, RevOps rebuilds, onboarding experiments. I’ve seen where things break. I build the systems that don’t.

Focus now:
→ AI agents as GTM operators
→ Infra that sells and scales without headcount
→ Blending engineering logic with sales instincts

Highlights:
– $200K ARR in <3 months at CozmoX (YC-backed)
– 438% MRR growth at UserGuiding, drove 57%+ new revenue
– Outbound engine + reseller motion at Shiny, >5% CVR
– Rebuilt RevOps at AstroLabs, hit 100%+ quota
– +30% retention, +17% activation at Cura from onboarding flows

What am I passionate about? building clean systems and watching agents in runtime actually doing the work..

Head of Business Development

I’m happy to share that I’m starting a new position as Senior AI Agents Forward Deployed Engineer at Beam AI!

Building AI agents & automations l Beam AI l Kusp AI

The main rules of SaaS still apply to AI agents businesses.. well, mostly..

CLING to the client's problem. That never changes.

But here's what I keep seeing: everyone can build an n8n or LangFlow or whatever automation now. Or Claude Code and Replit for shipping fast.. Prompt chains, agent loops, tool calling - it's all a weekend project. 
The build is commoditized.

So where does value come from?
Not from what you can build anymore. It comes from how long you've been in the trenches with a specific problem, and for whom exactly. 

The edge cases you've already hit. The guardrails you learned the hard way. The deployment speed you've earned through repetition.

AI agents actually open up more room for verticalization than traditional SaaS ever did. 

You're not just automating a workflow - you're encoding JUDGMENT. 
Domain expertise. The stuff that used to live only in people's heads.
That's a deeper wedge. 
And it's harder to copy..

But here's the uncomfortable part: finding that niche - the one that generates real, sustainable ROI - is genuinely harder than SaaS. 

The trust threshold is higher. Clients aren't just buying software. They're betting you won't automate them into a compliance disaster.

Once you earn that trust though? You start layering. SaaS features on top of your agentic core. That's how you expand.

The playbook for 2026 feels clear to me: unless you're already playing infrastructure with serious resources, your leverage is the wedge. 
One use case. Painfully specific. Relentlessly deep..

The generalists will demo well..
The specialists will get paid..

#aiagents #agentic #ai #saas

I believe we're not competing on better processes anymore. We're competing on better context. That's the actual game..

Last decade was about data integration. The edgs was with cleaner pipelines, unified stacks, faster decisions xx. That changed..
We are now playing the context integration game. 

Building a company brain that everyone can talk to. 
Your sales team asks it questions. Your engineers query it for decisions. 

It's the collective narrative- the ideation partner that actually knows your history.

The infrastructure's forming to achieve that but still immature. 
We have now some scattered components, MCPs, RAG which is getting sharper and models scaling context windows.

Nevertheless, the reality is still not as glamorous..

MCP is expensive. It standardizes cross-tool context, but integrating MCP context expands input tokens by 236× on average.
In fact, tool descriptions alone consume massive context window space, and with thousands of tools connected, agents process hundreds of thousands of tokens before even reading a request.

As for LLMs, the bandwidth is costly- 1MB of output costs around $1 per request, and you're billed in every follow-up message that includes that tool result..

At scale, I've seen this get prohibitive fast..
RAG is still fighting fundamentals. Precision and recall aren't solved- chunks misalign, relevant info gets missed, data goes stale..

But I think the vision is right..
This is about creating a fresher brain you can revoke later..

How do we speak better in meetings? 
Think more structurally? 
Document so context doesn't die in Slack? 

I believe the companies that answer this- that build the memory layer right- get an asymmetric edge..

It's not there yet. The tooling is immature. But the shift is real: 
from "do we have the data?"
to 
"can we surface the right context when it matters?"

My advise to anyone building:

Don't wait for perfect tooling. Start documenting with retrieval in mind- structured notes, tagged decisions, searchable context. 
And budget for tokens like you budget for compute..
The companies that treat context as infrastructure, not a feature, will be impossible to catch in 3 years.. 

#ai #aiagents #llms #ml #b2b

taplio

Anas Duksi's Linkedin Analytics

Anas Duksi

Performance ScoreAnas Duksi's LinkedIn Performance

Anas Duksi's Best Posts (last 30 days)

Ready to grow your LinkedIn presence?

taplio

.css-6ytdur{background-image:linear-gradient(284deg, rgba(39, 43, 72, 1) 0%, rgba(33, 47, 183, 1) 100%);color:transparent;-webkit-background-clip:text;background-clip:text;}Anas Duksi's Linkedin Analytics

Anas Duksi

Performance Score.css-goub6s{font-size:var(--chakra-fontSizes-md);font-weight:500;color:var(--chakra-colors-gray-600);display:block;margin-top:var(--chakra-space-1);}Anas Duksi's LinkedIn Performance

Anas Duksi's Best Posts .css-1emoo84{font-size:var(--chakra-fontSizes-md);font-weight:400;color:var(--chakra-colors-gray-600);}(last 30 days)

Ready to grow your LinkedIn presence?

Anas Duksi's Linkedin Analytics

Performance ScoreAnas Duksi's LinkedIn Performance

Anas Duksi's Best Posts (last 30 days)