Why your AI strategy isn’t working

Telco execs, I know you’re working on adopting artificial intelligence (AI). Let me guess: you have an AI task force, you hired a Chief AI Officer, and you’re methodically evaluating vendors through a 12-month RFP process. Yet, Amazon’s AI-powered Amazon Go stores have eliminated checkout lines entirely. Klarna’s AI assistant now resolves two-thirds of customer service chats while cutting resolution time by 82% and driving $40 million in projected profit improvements. And you’re still debating which AI models are “safe enough” for your employees to use on a daily basis.

This isn’t happening because you lack resources or talent. It’s happening because telcos are trapped in analysis paralysis, applying traditional procurement thinking to revolutionary technology.

Stop treating AI like it’s mature technology

The telco industry is approaching AI like it has every other technology—but AI isn’t like anything we’ve seen in our lifetimes. When you bought your last billing system, vendor capabilities were stable. When you migrated to cloud, Kubernetes didn’t fundamentally change between RFP and deployment.

AI is different. New models drop weekly, costs are collapsing (99.7% in two years), and possibilities change constantly. By the time you finish your 18-month vendor evaluation, foundational models will have evolved three generations.

So what’s a telco to do? Here are 3 things you should STOP and 3 things you should START to get your AI strategy off the ground.

3 things to STOP doing with AI

1. Say no to RFPs. By the time vendors respond to your ridiculous RFP, the approach they’re proposing has changed—twice. Anything written in an RFP will be obsolete in a month, so it won’t help you select the best vendor.

2. Don’t be a lemming. No one has mastered AI implementation at scale in telco yet. So you’re not going to find reference accounts where vendors can share their success with AI. That’s because telcos only have small use cases running in production. Instead, be bold and go first. Look for vendors who are walking the talk, those who can build deployable use cases quickly and are using AI internally themselves with a talented team that really knows AI. (Like Totogi! Full disclosure: I’m the acting CEO.)

3. Ditch the pilots. By the time you’re ready to scale your six-month pilot, the AI landscape will have shifted again and your tech will be obsolete. Instead, put it out to real users in your ranks to start building their experience working with AI.

3 things to START doing

1. Find real pain. Pick a real problem to solve—support tickets, escalations, neglected areas that are costing money—and tackle the problem with AI. Start with something so easy, it’s almost a negligible impact if it’s fixed, but that will give you a solution you can build on.

2. Build in public. Once you have something working, get real users banging on it. You’ll learn what’s not working, if it’s producing quality answers or not, and where the holes are in your solution. AI is not useful if it’s always in a demo state, or in the developer’s environment only. Get it out into the world, and get feedback.

3. Iterate daily. Sit your AI engineers next to the people using your AI tool, so they can watch real workers use it. They’ll find gaps, see new use cases, and improve it immediately. When your workers see AI tools eliminating mundane tasks, they understand the benefits and get excited, which creates a productivity flywheel. Once things are cranking, expand AI initiatives to more products, users, and processes.

Congratulations! You now have AI in production.

AI in the real world

A Tier 1, Asian telco CloudSense customer recently selected Totogi’s BSS Magic to automate enterprise sales tasks. We started small—sales for one product line—and aimed to improve the current process, where expert humans took 5 minutes and 50 clicks. Our first task was to see how fast AI could do it with no humans in the loop.

In two weeks we built an agentic flow that reduced the process to 50 seconds—a 94% time reduction. What used to require scarce product experts can now be done by an AI agent. Revenue bottlenecks disappeared overnight. Now the telco wants to expand to more products and more complicated use cases.

That’s how you do AI.

AI is your competitive advantage—if you act now

AI can eliminate hours of manual work, accelerate decisions, and unlock revenue streams. But every month you spend in committee meetings is a month your costs stay high while revenue growth stays flat.

Stop trying to moonshot AI. It’s something you build one win at a time. So start winning!

Recent Posts

Get my FREE insider newsletter, delivered every two weeks, with curated content to help telco execs across the globe move to the public cloud.