There's opportunity to differentiate on commsat technologies like switching and antennas, on offering different targets for latency/bandwidth vs price, planet-side downlink stations, and coverage at different latitudes: In particular, satellite-to-smartphone links, specifically optimized for lower-bandwidth, higher-latency, lower-power, lower-cost long-tail coverage. Fly a little higher, so you don't have to refuel as often, customers just trying to send a text can tolerate a little latency.
The Starlink terminal is great for grid-tied residential/RV/ship-sized home internet. It runs in the Ku and Ka bands at something like 65 watts. It's astonishing that they can get some functionality with their T-mobile partnership, but for the most part, it's not optimized for that.
What fraction of Starlink (or, hypothetically, AST SpaceMobile) expenditures is satellite launch cost per kg? The ATT market cap is $120B, if they can partner with/aquihire a $675M satellite company to stay competitive then even a $2.4B extra expense is not a huge deal. Sure, Starlink could do the same, but ATT has a lot of inertia. SpaceX won't knock them out of the market just by flying another set of satellites a little higher with some lower-bandwidth antennas.
If I could get voice and text anywhere on the planet (namely, in the woods in northern Michigan, population density ~1 person per square mile, or offshore on Lake Michigan, population density ~0 per square mile), that would be a game changer. Specifically, everyone knows that the cell networks lie about their coverage maps, this would let them genuinely say they have coverage everywhere. I'm currently using Garmin InReach for off-grid connectivity, not sure I'd dump that hardware and network for an esoteric feature that runs on my less-reliable glass slab, but a lot more people have glass slabs than InReach.