Hashim Syed

Early life and background

Syed was born in London and grew up there. He lived in the Netherlands for a large part of his childhood. He is originally from Pakistan and has also spent time in the Middle East. He has deep cultural roots in Pakistan and went back and forth between Pakistan and Europe. [1][2]

Education

Syed studied PPE, which covered politics, philosophy and economics, at university. He says he is not technical. He did not expect to work in AI while studying PPE. He later moved from studying philosophy and economics to leading AI strategy at Google. [1][2]

Career

While still a student, Syed started his own consulting firm, after growing frustrated with endless job applications. He has said that mentors and access to industry thought leaders led to his first internship, at EY, in asset wealth management. He also spent 10 weeks at JP Morgan and then decided to pivot, because he liked sales but not fixed income. [1]

Syed says he went to Meta while building his own company through viral distribution and the power of consumer intent, and wanted to learn more about this. He spent time at Meta and then started the UK's first VC startup growth program. Wanting to build again, he became a founding member and co-founder of a health and fitness app. [1][2]

Syed went back into big tech at Microsoft, leading a sales team for financial services. He became an entrepreneur in residence at Antler, saying he again had the itch to build. He has worked continuously for about eight years since his bachelor's degree. [1][2]

Google

Syed is Google's AI GTM Lead for North America. He has worked at Google for about four and a half years. At Google he first led product strategy for Search in the UK, starting in London. He was then selected for a special project in San Francisco. [1][2]

He helped start a Google team that does go-to-market work for DeepMind. He says his team is helping some of the most ambitious founders build and scale on Gemini, and that he works with companies from seed to Series A. He says his current job did not exist three years ago and that he went and found it. He is based in New York but travels a lot, and he says he loves what he does and considers it one of the greatest privileges to be in the unique position of working in a model lab. [1][2]

Work with AI tools

As a nontechnical person, Syed has said he was asking what coding is and what an agent is at the start of 2025. He has said he did not have any friends when he moved to San Francisco, and that he built an AI agent for scheduling with the sole purpose of being able to go to hackathons. He says he also builds his own AI tools. He uses a team of AI agents for his work but still meets clients himself, and he says he has trained an AI agent with his voice and face, but it is still not him. [1][2]

Views on AI and work

Syed says AI-related roles grew by about a thousand percent year over year and that AI labs are hiring for storytellers and content managers. He compares the AI shift to the internet age, when social media and digital marketing jobs appeared, and says new AI roles are creating a whole new industry. He is more optimistic than doom-minded about the job changes AI will bring to the industry. [1][2]

He says human skills such as empathy, critical thinking and leadership are durable, and that people need to focus on human skills and turbocharge them with AI-native skills. He says you have to be highly adaptable. He says people can outsource intelligence to AI but cannot outsource understanding. [1][2]

His advice to newcomers is to talk to an AI model, describe their work and ask it to teach them. He says he learned AI this way as a non-technical person. He says anything an AI model says should be questioned, as he was taught in philosophy. He says people should be curious learners about AI in this window before AGI arrives. [1][2]

Views on startups and enterprise AI

Syed says companies once had five to seven years to reach 100 million in annual recurring revenue, while in the AI-native environment tools such as Cursor reach that same scale in five to seven months. He borrows Bessemer Venture Partners' idea of supernova companies that grow to 40 million ARR in a year, and describes typical SaaS companies as shooting stars. He says there is no rule for how fast a company must grow, and he learned this as a founder. He says a startup does not need a minimal viable product but a minimum lovable product. [1][2]

He has said he did not want to burn lots of money on advertising and went to market by turning up at gyms and giving out flyers. He says efficiency is now table stakes, and that every startup and enterprise company is focused on AI. He says the question is how to re-engineer and redesign end-to-end processes, a subject he says can be discussed through agent orchestrations. He asks how AI can do more than baseline productivity and deliver new areas of growth. [1][2]

He says governance and education are the enablers for scaling go-to-market with AI, and defines shadow AI as staff using AI tools in unlicensed ways at work. He says enterprise risk management matters for both large and small companies, and that data inside Google's enterprise platform belongs to the customer, who owns and controls it. [1][2]

He says token cost matters to customers at every stage, from seed startups to large enterprises, and describes token cost as somewhat the bottleneck for agentic systems. He says peak frontier models can be too expensive for sub-agent tasks. He compares AI agent teams to human teams, with a manager and specialists underneath, and says the model that runs the team's brain differs from the model that serves individual sub-agents. He says the aim is to align the right model to the right workflow and the right user. [1][2]

He says 2025 and 2026 are the years of vertical AI, with agents built for specific industries. He says real value is usually customized to many different workflows, and that the real winner will be the founder who builds the best workflow integration and product experience on top of models. He says AI systems should be seen as collaborative partners rather than scary systems. [1][2]

Views on models and technology

Syed describes Gemma as Google's open-source model within the Google ecosystem, and says it is cheap and fast and that developers can host and fine-tune it themselves. He says open source and closed source are not a zero-sum choice, and the world needs both. He says the real question is the right model for each use case, not closed versus open. He says Google's strategy is to be the hyperscaler that offers both closed and open models, that its platform is interoperable so builders can use whatever model they choose, and that Google wants to be the hyperscaler for the entire stack. He says Google is one of Anthropic's biggest partners and has trained and run Anthropic on GPUs. [1][2]

He says Gemini has always operated in the middle of the frontier of price to intelligence, and that it is one of the best models on the market for multimodality. He says Gemini Live is available in most languages around the world and that AI needs localising to different markets, cultures and languages. He says Gemini Enterprise is live in 90% of Fortune 100 companies, and that a range of startups use Gemini and Google for Startups credits. He recommends Google AI Studio for prototyping. [1][2]

He says much of the world is stuck without the foundational data layer AI needs, not because of model access. He says scaling up text-predicting LLMs will soon reach its peak, that world models are the next step because they understand physics and spatial logic, and that deep-thinking agents are given more time and compute to test options before answering. He says AlphaFold is a real breakthrough that solved 3D protein structure prediction. [1]

He sees physical AI and robotics as the new frontier in AI, and expects humanoid robots and autonomous vehicles to be more common in daily life. He says the world will look very different in five years and that we are in a precious window before AGI arrives. [2]

Pakistan

Syed says Pakistan is the world's third largest exporter of IT freelancers, and argues it is better placed than non-technical markets to become AI native. He says Google has worked with many Pakistani companies on Gemini Enterprise training and licences, and is working with companies in Pakistan on safe and secure AI in the workplace. He says Google has led a fellowship with Googlers from around the world, including Singapore teams, working with companies inside Pakistan. He says Google recently ran a nationwide coding competition focused on AI enablement and trained almost up to 100,000 people through AI. [1][2]

Future plans

Syed says he expects to return to building in the near future and hopes to return to the founder ecosystem as things calm down. He has said he is talking to a lot of VCs and investors these days. He has ideas in physical AI that he would like to explore. [1]

Timeline

2025

Asked what vibe coding and an AI agent were, at the start of the year[1]

Appearances

Frequently asked questions

References

1. You can outsource intelligence, but not understanding . Beyond the Bubble , 2026-08-12.
2. You can outsource intelligence, not understanding . Thought Behind Things , 2026-08-07.