Hello People »
The architecture of action
At the intersection of AI, enterprise systems and human judgment, Murtuza Haryanawala is building technology that turns intelligence into execution
By Team Vibe
Meet Murtuza (Matt) Haryanawala, a visionary tech leader and startup strategist bridging the gap between complex engineering and human connection in the San Francisco Bay Area. With a remarkable track record of scaling partner solutions past the $1 billion ARR mark, including spearheading Broadcom’s groundbreaking Wi-Fi chipsets for Samsung, Murtuza thrives at the intersection of strategy, technology, and community. Today, he leads his startup Tixa, building on his cutting-edge Gen AI initiatives at Productlabs.us, and his community leadership via Llamalabs.us, he brings a unique, fun-loving human spirit t building AI.
Excerpts from our conversation:
Every AI founder starts with a problem that drove them nuts. What was the single most painful customer support or operations meltdown you witnessed that made you say, “We need to build Tixa right now”?
A friend who ran a manufacturing business in the Kitchen and Bath industry out of Dallas TX, mentioned how customer support was a critical part of his business operations. He shared how his team of 10 human agents worked round the clock to service 300 plus daily tickets. To resolve a ticket, a typical agent had to open five different tabs: the CRM, payment gateway, core banking engine, compliance logger, and ticketing system. They spent 20 minutes copying and pasting, tracking IDs across systems just to make a two-second decision. His challenge was to improve CSAT while preventing agent burnout. The issue wasn’t a lack of knowledge but a breakdown in execution. I realized then that support teams didn’t need another knowledge base tool to draft prettier email replies; they needed an intelligent execution layer to actually do the work.
Most people hear AI support and think of annoying pop-up chatbots that never help. Tixa promises to resolve work rather than just chat. How do you bridge the gap between AI just talking and AI actually taking action?
The difference between chatting and executing comes down to system integration and trust architectures. Chatbots fail because they operate purely on text generation — they read an FAQ and regurgitate an answer. Bridging the gap requires treating the AI as an agentic system with direct, permissioned access to business context and intelligence via structured protocols. Instead of the AI saying, “Here is our policy on address changes,” Tixa validates the user’s identity, checks order status rules, calls the logistics API, updates the shipping provider, and confirms the change in the database. Execution means moving from passive conversational UI to active, deterministic workflow completion.
In customer operations, what is one wildly back-office task that Tixa automated that made support teams secretly jump for joy?
Getting access to contextual business data and cross-system reconciliations, that was tedious, error-prone, and soul-crushing. Automating that entire back-end lookup, verification, and database sync—so the ticket resolves itself in seconds— is the “aha” moment for teams.
There’s a lot of fear around AI taking human jobs. What’s your opinion?
AI isn’t eliminating human capability; it’s eliminating human toil. When AI takes over the repetitive execution layer, humans get elevated to high-empathy edge cases, complex dispute arbitrations, and proactive relationship building. The future isn’t zero humans; it’s humans operating at the top of their intelligence, backed by an autonomous digital workforce. As we deploy, we learn AI can do 80% of a job, but a human still needs to do the 20%. This means the productivity gain for a human is now 5X, allowing a business to grow more and faster, using the same human talent pool.
In fintech and regulated operations, a single hallucination or mistaken transaction can ruin trust. How do you build “guardrails” so AI never accidentally promises something it shouldn’t?
Trust isn’t built on emotion or hopeful prompts; it’s built on architecture. In regulated environments, you cannot treat an LLM as an unconstrained decision-maker. We build a strict “Trust Stack” around the AI. This means separating reasoning from execution: the AI uses deterministic guardrails, structured API schemas, and hard business rules before any transactional action is executed. If a request falls outside verifiable parameters or triggers an anomaly threshold, the system automatically falls back to a human-in-the-loop workflow with full context attached. You never let a probabilistic model make an unvalidated, high-stakes decision.
What’s the biggest difference between support expectations in fast-moving markets like India versus global markets in the US or Europe?
In fast-moving emerging markets like India, speed, hyper-responsiveness, and channel preference dominate. Customers expect instant resolution directly on conversational platforms like WhatsApp. They don’t want to wait 24 hours for a polished email. In US and European markets, while speed matters, there is a much heavier emphasis on strict data privacy, formal compliance frameworks, and structured step-by-step clarity. India pushes you to scale for massive volume and real-time execution; Western markets push you to scale for deep governance and policy precision.
What is one current AI trend or buzzword that you think is complete marketing fluff?
“Wrapper” chatbots branding themselves as “Autonomous AI Agents.” Simply taking a basic prompt, adding a search tool, and slapping an “Autonomous Agent” label on it doesn’t mean it can handle real enterprise operations. Real agency requires transactional integrity, multi-step system orchestration, robust error recovery, and enterprise-grade permissioning. Most of what gets sold as “agentic” today is just conversational text generation with a fancy wrapper.
Beyond metrics, growth rates, and product roadmaps, every founder has a core set of non-negotiables. What is one personal moral or core belief you refuse to compromise on?
Establishing trust at every layer in an AI stack is critical. This also includes the human layer and uncompromising integrity in how we build and present trust. In the tech ecosystem, it’s easy to over-promise, hide system flaws behind clever UI, or cut corners on data handling just to close a deal faster. I refuse to operate that way. Whether it’s the security of our data architecture, clear communication with customers, or how we treat our own team, transparent, verifiable trust is non-negotiable. If a product feature isn’t production-ready or completely secure, we don’t ship it to hit a vanity deadline.
Tell us about a specific moment in your journey where things went completely sideways. What did that experience teach you about your own character under stress, and how did you treat the people around you in that moment?
A critical junction in our journey was our design partner’s decision to develop an adjacent product for their industry. This meant we needed to go back to the drawing board, and rebuild our entire GTM strategy from scratch. The big lesson was to act as a shock absorber for the team and protect the team’s morale and directional clarity. This setback helped build resilience and team culture far more than the wins we had celebrated.
If you could force every company founder to adopt one golden rule for handling customer support, what would it be?
Treat customer support as an opportunity to create delightful customer experiences that lead to retention and growth. Stop treating customer support as a cost center, and start viewing it as your product’s real-time truth engine. It is an honest opportunity to fix underlying product bottlenecks permanently.
If you were stuck in an elevator with a skeptical COO who says “AI support agents just confuse my customers,” what’s your quick, mic-drop response to change their mind?
Chatbots dressed as AI support agents confuse customers because they only give FAQ type answers. Customer support is becoming an operational intelligence layer, not just a service function. The win is real-time intelligence in the moment of customer interaction to resolve and complete a workflow.