The business wants AI, and it wants it yesterday. But once the proof of concept works and the partner goes home, someone inside the company has to keep the thing running. Nathan Hill, Head of Telco at AWS and a 20-year veteran of the telco industry, joins James to talk about what actually happens when enterprises push AI into production — and why the scarcest asset in the room is still deep domain knowledge. In this episode: - Why "we need to do AI, the board's pushing for it" so often collides with the buy-versus-build question no one has answered internally. - Where AI rollouts go wrong: over-engineering a basic problem, or buying an off-the-shelf tool and expecting it to be bespoke. - The proof-of-concept trap — projects that "prove AI works" but were never built with a path to production. - Why "AI native" doesn't translate cleanly to banks and telcos carrying 20–30 years of legacy and technical debt, and why the outcome should drive the tech strategy, not the reverse. - The handover problem in one line: "If your chatbot starts spitting out Gordon Ramsay recipes instead of the answer, who in your organisation can fix it?" - Centralise-then-seed: standing up an AI centre of excellence without creating an isolated team of "cool kids" cut off from the business (with a NASA analogy on risk). - Why you can't take 20 years of experience, grab a dev and say "now you know telco" — and how the tooling finally lets domain experts build. - Human-in-the-loop for critical infrastructure, digital twins of telco networks, and ICs becoming managers of agents. - The question James keeps asking: if I'm managing a team of agents at 10x productivity, what should I actually be paid? - Deep versus broad careers, how go-to-market has changed, and whether it's easier to teach a salesperson the tech or an engineer to sell. Nathan Hill is Head of Telco at AWS, leading the company's engagement across Australian telcos. Before AWS he spent more than 20 years inside the telco industry — operations, engineering, pre-sales and sales — and ran sales and marketing for a challenger telco. He has a rare view across both the technical build and the go-to-market that sells it. Building the team that has to make AI stick in production? This one's for you. Connect with Nathan Hill on LinkedIn. Learn more about AWS in telco at aws.amazon.com. --- Episode Summary The business wants AI now — but who keeps it running once it's live? Nathan Hill, Head of Telco at AWS and a 20-year telco veteran, sits down with James MacDonald to unpack what really happens when enterprises move AI from proof of concept to production. They dig into the buy-versus-build decision most organisations skip, why so many AI projects stall with no path to production, and the operating-model questions — who maintains it, who fixes it when it breaks — that companies leave until it's too late. Nathan makes the case that deep domain expertise is the asset AI can't replace: you can't grab a dev and say "now you know telco," but you can finally give 20-year network engineers the tools to build. They also get into human-in-the-loop for critical infrastructure, ICs becoming managers of agents and what that's worth, deep-versus-broad careers, and how the go-to-market function is changing now that the salesperson has to understand the tech. Practical, grounded, no hype. Time Stamps 0:00 "I'm managing a team of agents now — what should I be paid?" 1:30 Meet Nathan Hill, Head of Telco at AWS 2:58 "The board wants AI": the buy-versus-build question 4:30 Where AI rollouts go wrong 6:35 Start with a use case — but build a path to production 8:09 Is "AI native" realistic for banks and telcos? 9:36 Speed versus security in regulated industries 11:56 The handover problem: who maintains it? 15:21 AI centres of excellence and the NASA analogy 18:26 "You can't grab a dev and say now you know telco" 21:30 Why agents won't replace engineers 23:41 Managing a team of agents — what's that worth? 25:03 Deep versus broad: staying in one vertical 27:08 How go-to-market has changed 29:59 Teach a salesperson the tech, or an engineer to sell? 32:58 Building a go-to-market function from scratch 38:03 Why the "SaaS apocalypse" is wrong 39:56 Upskilling into modern go-to-market 41:21 Career advice: back yourself 42:42 James's takeaway: domain experts who learn to build About the host James MacDonald is the founder and Managing Director of NTP Talent (Newy Tech People), an Australian tech and engineering recruitment firm headquartered in Newcastle with teams in Sydney and Melbourne. He hosts Building Tech Teams, helping companies up the East Coast of Australia find and recruit the best technology talent. Connect with James on LinkedIn (/JamesMacDonaldAU) or at ntp-talent.com.au. About Day One Network Day One is a podcast production company and trusted partner in the technology space, producing shows for founders, investors and operators across Australia and beyond. Building Tech Teams is part of the Day One Network, which cross-promotes episodes across a slate of technology and venture shows. Building Tech Teams is produced by Day One®, trusted partners in the technology space and the production partner behind Blackbird Ventures' Wild Hearts. Sister shows include First Cheque, Oversubscribed and In The Blink of AI. Episodes are cross-promoted across the network.
