Most “AI for sales” and “AI for recruiting” products stop at drafts, rankings, and dashboards. The hard part is the last mile: reaching a human, holding a compliant conversation, extracting evidence, and closing a loop into a calendar or CRM.

Nexora attacks that last mile for outbound revenue — voice agents that behave like operators, with telephony orchestration that actually gets through voicemail, screeners, and IVRs. Hurraay points the same engine at recruiting’s new flood: AI-written applications that still need human-quality screening calls, ranked shortlists, and interviews booked in days — not weeks.

The company is worth existing because it owns a scarce asset: production conversational infrastructure in Europe, plus two vertical wedges where the economic outcome is obvious (pipeline and hires). Shared platform, two pains, one compounding learning loop from every call.

That is my working thesis. As founding PM I would pressure-test it weekly against real customers, real transcripts, and real conversion — and kill whatever part of it is wrong.

Shared voice substrate

Hurraay is new; the engine underneath is not. Thousands of scheduled, multilingual, compliant calls imply a production telephony + voice-agent layer: turn-taking, barge-in, latency budgets, recording consent, EU data residency, and audit trails. Nexora’s public agent map — Campaign, Voice, Booking, Integration, Call Scheduling, Success Improvement — reads like an outbound operating system, not a single prompt.

Closed-loop agents, not chatbots

The valuable system is the loop: goal → script/behavior → call → transcript evidence → score or CRM update → reschedule or book → A/B learnings back into the next campaign. Hurraay’s recruiting flow (criteria agreed up front → résumé score → live screen → transcript-linked rank → calendar book) is the same loop with a hiring-manager shortlist as the payoff.

Human steering as a product feature

GDPR Article 22 and “no machine rejections” are not footnotes — they are product constraints that force evidence-linked scores, parked candidates instead of silent rejects, and one-click human override. That is hard to bolt on later; it has to live in the evaluation UX.

What I would dig into first as PM

1. Where conversion dies: connect rate, consent, script drift, booking no-shows.
2. What “success improvement” actually measures and how fast learnings ship.
3. How much of Hurraay vs Nexora is shared code vs shared brand vs shared ops.
4. The minimum enterprise sales workflow after “book a meeting” that expands ACV.

Business vision: become the default conversational revenue (and hiring) engine for Europe — self-serve enough to move fast, serious enough for enterprise compliance, and disciplined enough that no single customer owns the P&L.

• Near term: prove repeatable unit economics on Hurraay’s pay-by-hire and Nexora’s outbound seats/campaigns in DACH.
• Two years: ~€10M ARR with a healthy mix across products and accounts — no customer over 20% of revenue.
• Five years: ~€100M ARR as the platform expands use cases while vertical products keep distribution sharp.

The strategic bet is platform leverage: every outbound call and every screening call improves connect logic, conversation quality, and booking reliability for both products. Founding PM work is choosing which wedge funds the platform next — and saying no to shiny agents that do not move revenue or time-to-hire.

Product vision in one line: agents that complete revenue and hiring workflows end to end, with humans setting criteria and owning irreversible decisions.

What “next” means to me

1. Deeper enterprise sales paths — beyond first-meeting booking into renewal, reactivation, qualification, and multi-step outbound that CRM teams trust.
2. Recruiting as a reference wedge — Hurraay shows the market what evidence-backed voice screening looks like when humans stay in control.
3. Evaluation as the moat — transcript-linked scores, comparable criteria, and manager-readable reports beat “vibes” copilots.
4. Self-serve without losing craft — minutes to first call, without becoming a toy that enterprises cannot audit.

I would sit on real calls, read failed transcripts like bug reports, and write specs that engineering can ship in a week — not quarters.

Operating cadence

• Week-by-week shipping. One customer-visible improvement per week beats a perfect roadmap.
• Argue from evidence. Transcripts, connect rates, and hire/book outcomes settle debates louder than opinions.
• PM + eng as one unit. Specs are short, live next to the code, and change when the call data changes.
• Radically honest reviews. Kill features that do not move the North Star — time-to-qualified-conversation or time-to-interview.

How I show up

I will not wait to be told what to build. I will bring a point of view, pressure-test it with customers and the founders, and pull in mentors when I am stuck. That is the process commitment behind everything else on this site.

I have shipped software that had to work in messy, high-stakes systems: incentive fraud on €890M annual spend and maps/location economics at Delivery Hero; earlier fintech and ride-hailing product work at Careem and Ola. That is the “shipped at scale” muscle your JD asks for.

Separately, I am rebuilding as an AI-native PM: shipping RAG, MCP, and agent projects, writing Sorted Signals, and running a listening tour with senior product leaders — including conversations with Mohammed Azzam (Director of Product, Parloa) and Rupesh (CPO, BuddyWise).

What makes me unusual for a founding AI PM role is the combination: operator depth from large platforms, plus the willingness to start without a brief, argue for the hard product calls, and learn in public. The next page is the honest version of how I learned that the expensive way.

Read the Securifi history →

Coming soon.

Three failures at Securifi

1. I waited for the CEO to tell me what to do. Early-stage does not hand you a roadmap. I treated ambiguity as someone else’s job.
2. I was not bold enough. I had views on what we had to build and what we were getting wrong — and I soft-pedaled them instead of fighting for the product.
3. I did not use my network. I tried to invent the job alone instead of asking people who had already done it.

Ten years later — three corrections

1. Self-starter. I run my own project, Sorted Signals, with a monetization plan. I do not wait for permission to ship.
2. Courage as practice. Boldness was my resolution for 2024 — and I kept it. I resigned from Delivery Hero when the path was going nowhere, without an offer in hand. I have been creating my own work in the 2026 market and reaching out cold to leaders across the industry.
3. I ask for help on purpose. I have built a mentor network to get better at this job — including Mohammed Azzam (Director of Product, Parloa) and Rupesh (CPO, BuddyWise).

Founding-stage AI product work needs someone who starts without a brief, argues for the hard product calls, and pulls signal from customers and mentors fast. I learned those the expensive way once. I will not learn them again on your runway.

Those numbers are a shared scoreboard, not a personal vanity metric. They force product choices that expand the market, deepen platform leverage, and refuse dangerous logo dependence.

If that ambition matches how you want a founding PM to show up — start here, argue loudly, ship what matters — I would like to talk.

Email me LinkedIn

Accept reality and succeed

LinkedIn EasyApply stopped working in 2019. I sent out 150 applications and got 1 interview. Now I’m turning up my LinkedIn presence, attending events, and building a network. Now I’m getting inbound interview invites.

I accept reality, and expect the same from everyone around me.

I can jump out of a plane, but I’ll need a parachute. Reality doesn’t forgive anyone!

Set high standards

Robin asked for an email. I created this webpage instead. I believe in standards, even for the small things.

Give credit where credit is due

I believe in crediting people for their contributions, no matter how small. Having someone steal credit for their work is a huge demotivator that trains teams to stop contributing.

I credited Robin for the idea he gave me about marketplaces.

Just enough process

You don’t need daily stand-ups (I hope you’re not doing them). We do need weekly idea sessions (see Process Suggestion). Process should serve us. Not the other way around.

High Time-To-Aha!

The time it takes to demonstrate compelling value to prospects on the Nexora site is well over 3 minutes.

I signed up for the demo account. There was a LinkedIn sign up screen, then an unskippable guided tour, a lengthy form filling process to create the first campaign… which didn’t work. Also, you currently do not verify the phone number with an SMS. That means I can enter any phone number that I want and make at least one call to it.

De-risk the demonstration of value by switching to a “Try It” button instead which takes just one input, a valid phone number, then verifies it with an SMS first. Make a call to the prospect immediately upon successful verification. The prospect experiences the product much faster. Copy along the lines of “You received the call, now send calls out” to guide them to sign up.

The expected outcome is higher rates of activation. The #1 risk is higher costs from telephony APIs, Eleven Labs, and LLMs from making calls to many unqualified leads.

I would manage the risk by using API cost per converted lead as a guardrail metric, and lift in activation rate as the objective metric for this idea.

Marketplace Pivots

Robin’s idea: A pre-vetted database of qualified professionals companies can take directly to hiring manager rounds. Hurraay conducts the equivalent of recruiter screening via voice agent calls.

My add-on: A skill assessment marketplace, or partnerships with Maven etc. that can add credibility to job candidates on Hurraay.

Interview Practice

Attract job candidates to Hurraay by building an interview practice side to the product. Adds another source of revenue, and users who sign up for interview practice can join the roster shown to hiring teams with one click.

This is already a proven idea (benerez.com/copilot), but not widespread outside Silicon Valley. Hurraay can be the first, or very early, in the European market.

It also strengthens the existing business of boosting signal and introducing reliable candidates to employers.

Job seekers should be able to consent to using their practice rounds to inform their candidate scores. Off by default, privacy first.

Moats

Eleven Labs, Twillio, Claude Haiku. These are available to everyone.

Hurraay / Nexora have to focus on generating new data from operations. Proprietary data that no other frontier lab, AI shop, or old school SaaS has makes the business less easy to replicate.

Weekly Idea Session

Gather the whole team every Saturday morning. Everyone shares their new ideas for growing the business. Every idea is welcome, no one is discouraged, and credit is assigned publicly and permanently. Tie non-extrinsic rewards like a trophy, or an exclusive hat, to top contributors.

Async ‘stand-up’ by 6pm

No need of pulling everyone away from tasks, forcing a context switch, only to hear “no updates” or “I’m working on it”. Everyone submits what they’d like to say to an agent by 6pm. The agent summarizes and manages updates to leaders.