01
Diagnose
Deliverable
A workflow map — the process, the data, the roles, the exceptions, and the real cost of the manual version.
Company
AI systems for the workflows your business cannot afford to run manually.
Profitec AI designs and implements AI-enabled operating systems across workflows, CRM, reporting, knowledge, lead management, and internal operations — with clear controls, measurable outcomes, and integration into the tools your teams already use.
operating-architecture · referencelive
How we wire an operating system around a workflow: source systems are validated, an AI decision layer acts, and every result either writes back to your tools, updates a live dashboard, or escalates to a human-approval gate.
Who we are
We rebuild the operational core of a business — not just bolt on automations.
We design AI operating systems for workflows where fragmented data, slow decisions, and manual execution create measurable operational risk.
Profitec AI is an Israel-based AI systems and automation company for B2B. We combine engineering, business-process design, and AI implementation — because durable automation is not a tool you install, it is an operating contour you redesign around the data, the roles, and the decisions a business actually runs on.
We work across workflow automation, CRM, reporting, knowledge and RAG, lead management, document processing, and internal operations — for clients in finance, legal, healthcare, professional services, and operations-heavy B2B. Our background is in controlled AI decision systems: data pipelines, multi-agent workflows, risk gates, human approval, dashboards, and monitoring.
Company
- Legal entity
- Profitec Israeli Innovation Center Ltd
- Founded
- 2020
- Based
- Ramat Gan, Israel
- Serving
- Israel · North America · Europe · United Kingdom
- Focus
- AI systems, automation, RAG, and workflow engineering for B2B operations
- Sectors
- Finance · legal · healthcare · professional services · operations-heavy B2B
What makes Profitec different
Four things we do that most automation vendors skip.
We sell the ability to re-engineer how a business operates — not a bundle of disconnected automations.
01
Process before tooling
We don't start with n8n, Make, or an LLM. We start with the bottleneck — the data, the roles, the exceptions, and the real cost of the manual process — and only then decide what to build.
02
Engineering, not chatbot theatre
Integrations, validations, fallbacks, monitoring, permissions, audit trails, and human approval wherever the business needs control. A demo is not a system.
03
Systems that fit existing operations
CRM, documents, inboxes, ERP, reporting, APIs, internal databases — automation is built around the tools you already run, without forcing the company to rebuild itself around AI.
04
Measured operational outcomes
We tie success to numbers that move: processing time, response speed, conversion, error rate, cost per task, recovery rate, and adoption.
How we work
We run delivery as an engineering project, not an “AI install”.
Five stages, each with a concrete deliverable — so you can see the system being engineered, not just promised.
02
Design
Deliverable
A system architecture — source systems, validation, the AI decision layer, controls, and the integration points into your tools.
03
Build
Deliverable
The implementation — agents, integrations, and the write-backs into the CRM, inbox, and systems your teams already use.
04
Validate
Deliverable
QA, controls, and acceptance criteria — fallbacks, permissions, and human-approval gates tested before anything goes live.
05
Improve
Deliverable
A KPI review and ongoing optimization, measured against the outcomes we agreed up front.
Proof
Systems we have built, published end-to-end.
No invented logos or borrowed metrics — real engagements, each with the problem, the system, and the operational impact.
Market Intelligence Crew
- Problem
- Fragmented market signals and slow, manual analysis, with no controlled way to turn a read into a decision.
- System built
- An 8-agent CrewAI system with a 200+ feature ML brain and a 7-check risk gate — multi-agent orchestration with data validation and risk controls.
- Operational impact
- A controlled decision pipeline published end-to-end — including the cases where the correct, risk-gated output was “no”.
AutoClaim
- Problem
- High-volume vehicle-insurance claims processed manually, case by case, with no consistent routing or review.
- System built
- A 5-stage operations layer with 7 specialist agents: intake, document analysis, routing, and reminders — and one hard rule.
- Operational impact
- Every unclear case returns to a human. Routine throughput is automated; judgment and sensitive actions stay human-in-the-loop.
Fershteyn Law
- Problem
- A Brooklyn estate-planning firm was effectively invisible in AI-generated answers across the engines buyers now ask.
- System built
- Schema markup, entity clarity, structured evidence that matches the page, and prompt-level citation monitoring.
- Operational impact
- Citation rate across Gemini, Perplexity, and ChatGPT rose from 16.1% to 62% in 8 weeks — a 3.9× lift.
AI Avatar Content Engine
- Problem
- A regulated law firm needed a daily brand-video presence it could never sustain by filming — and every public claim is a compliance surface.
- System built
- A self-hosted pipeline: RAG-grounded scripting, one-tap approval, avatar generation, automated post-production, and distribution to nine platforms.
- Operational impact
- ~90 published assets a month from a single recording session, with advertising compliance enforced in code and founder time cut to ~5 minutes a day.
Each case is published with its architecture, controls, and the honest result. The Market Intelligence Crew is shown as an engineering example of controlled AI decision-making — not a financial promise.
View all case studiesWho builds it
The people who design the system and own the result.
You are not buying “AI” — you are buying the judgment of the people who design the data flow, decide the controls, and stand behind the outcome: strategy and client outcomes, solution architecture, automation and integrations, AI and data systems, and delivery with post-launch support.
Nisan Flekman
CEO & Founder
Business leader focused on AI consulting and automation strategy. Experience across finance, technology, and healthcare, guiding companies from planning through implementation.
LinkedIn
Moshe Ben-Shimol
Co-Founder & Business Development Director
Attorney and entrepreneur with expertise in commercial and civil law, AI regulation, and business development. He leads partnerships and market-entry initiatives, with entrepreneurial experience across real estate, healthcare, fintech, and food-tech.
LinkedIn
Pazit Flekman
R&D Consultant
Engineering leader and AI developer focused on large-scale solutions across cloud platforms. Builds high-performance teams from concept through full implementation.
LinkedIn
Jonathan Roumani
Senior AI Developer
AI Engineering Lead specializing in AI agents, voice AI, SaaS development, and business process automation. Jonathan transforms complex workflows into practical AI systems that improve efficiency, scalability, and operational performance.
LinkedIn
Julia Zhilin
AI Video Content Maker
Video content creator focused on AI-generated videos for business promotion and branding. Turns ideas, services, and messages into modern visual content that captures attention.
Vladimir Zhemerov
Senior Product Manager & AIO/GEO Specialist
Product manager focused on translating business problems into deployed AI systems. Owns the path from client discovery to live workflow — connecting stakeholder context with the technical build behind it.
LinkedInSecurity, controls & AI compliance
Controls are part of the design, not an afterthought.
For finance, legal, healthcare, and enterprise operations, how the system is controlled matters as much as what it automates.
AI automation should not create an uncontrolled decision layer inside the company.
Role-based access
Every agent and integration runs with scoped permissions — access follows the role, not the convenience.
Data minimisation
Systems read and retain only the data the task requires, nothing more.
Approval gates
Sensitive or high-risk actions wait for a human decision before they execute.
Audit trails
What the system did, when, and on what input is logged and reviewable.
Monitoring
Live observability on throughput, errors, and drift — so failure is caught, not discovered.
Fallbacks
When confidence is low or a service fails, the workflow degrades safely instead of guessing.
API & credential handling
Keys and secrets are scoped, stored, and rotated — never embedded in the workflow surface.
Responsible AI
An AI-compliance approach to claims, disclosure, and accountability, built into how the system is designed.
Start here
Tell us about one operational bottleneck.
Describe one repeated process — its tools, volume, and where it breaks. We will review whether it is suitable for an AI operating system and come back with a practical next step.