It has never been easier to build software.
I've spent the last year rebuilding Bannerbear from the ground up for V5, and a large chunk of that work was done with AI coding agents. Features that would have taken me a week in 2022 take an afternoon now. That felt great right up until I realised that every competitor I have, and every competitor I don't have yet, can do the exact same thing.
If you run a traditional SaaS company, this is the new reality. The thing you sold for the last five years, a set of features behind a login, is now something a motivated developer can clone over a weekend with a $200 subscription to Claude. Not all of it, not perfectly, but enough.
So this post is about what to do about it. It's a list of the ways I see SaaS companies competing / pivoting in the AI era. Some of these I'm doing myself at Bannerbear, some I'm watching other founders try, and one or two I'm not convinced by yet.
Competing on features is dead
Let's get this out of the way first, because everything else follows from it.
A basic way to compete in the trad SaaS era was: find a competitor, build everything they have plus a few things they don't, and start marketing it. Feature parity+1 worked as a strategy because features were expensive to build.
Now they are cheap to build.
Fun story, one of our competitors still markets themselves as saying "10x more features than Bannerbear". Maybe that was true a year ago, it's definitely not true now, and any motivated dev at any tech company can destroy the validity of a statement like that in a matter of days. It just doesn't make sense to make claims like that anymore.
So where do we go from here? Frankly I think half the industry is still figuring this out. Those who had the good timing to launch an AI-native product in the 2023-2025 era seem to be doing well. So what about the rest of us?
Here are a few ways I think SaaS founders can still compete / pivot in the AI era, without resorting to a list of infinite features.
1. Build the unsexy things nobody wants to build
There is a category of software that AI has made only marginally easier: the stuff that's hard because of operations, not code. Hosting. App deployment. Email delivery. Payments infrastructure. Anything where the difficulty is in running it reliably at scale for years, dealing with abuse, keeping IP reputation clean, being on call at 3am.
Nobody builds a transactional email service "in a weekend" with an AI agent, because the code was never the hard part. The hard part is the deliverability relationships, the compliance, the sheer boredom of keeping the thing up.
This is where I'd look if I was starting again today. Boring, operational, unglamorous. Developers will happily pay for it forever precisely because they do not want to do it themselves, and an agent can't do it for them either.
2. Build where you have distribution
This has always been true, but it's more true now. If two products have identical features, and they will, the one with the audience wins.
Distribution can mean a lot of things: an existing customer base, a newsletter, a community, a free tool that brings in traffic every day.
If you're a founder with an audience, your audience is now worth more than your codebase. Build for the people who already listen to you, and make your product the obvious thing to buy when they're ready or recommend to others. If you don't have an audience yet, that's the project.
3. Build for agents, not humans
One of the more interesting pivots I'm seeing is companies going completely agent-first. Instead of a dashboard with buttons, the product is a set of tools an AI agent can call. The human (almost) never logs in. They tell Claude or ChatGPT what they want and the agent does the work through your API/MCP.
We did a version of this with the Bannerbear MCP server. Someone can sit in their coding agent and say "generate 20 product videos from this spreadsheet using my template" and it just happens. No clicking, no tutorial.
I don't think humans stop using software entirely, at least not soon. But I do think an increasing amount of the work will be done by agents on behalf of humans, and if your product isn't reachable by an agent, it will slowly stop being reachable at all. Even internally, lately I find myself telling my team to "just chat with Bannerbear" rather than click around in the interface. It's actually way quicker much of the time.
4. Go where the tech isn't
There are entire industries that never got a proper software revolution the first time round. Trades, logistics, agriculture, local government, small manufacturing, anything where the current "system" is a spreadsheet and a phone, or some UI that was last updated in 1997. These businesses are real, they make money, and their tools are decades out of date.
In tech we compete with each other for the same developers and the same marketers, and now we all have the same AI tools too. Step outside that bubble and your technical advantage is enormous again. I've been exploring this space personally.
The catch is that these markets are harder to reach. You can't launch on Product Hunt and expect plumbers to show up. But that's the moat. Boring, real-world businesses that need updating for the AI era are, I think, one of the biggest opportunities of the next decade, and they're mostly being ignored by people who could build for them.
5. Pivot into hardware
I'm seeing more tech companies attempt this.
Software is now cheap to copy. Hardware is not. A physical product needs manufacturing, supply chains, certifications, inventory, support, and every one of those is a barrier a weekend project cannot cross. Pair a device with the software you already have and you've got something that's hard to replicate.
The risks are obvious. Hardware is capital-intensive, margins are thinner, and a software founder usually has no idea what they're doing in a factory. I'm watching this one with interest rather than doing it. But if you have the stomach for it, a hardware pivot has more of a moat than almost anything else on this list.
6. Lower your prices
This one is going to be unpopular. For years the standard advice, including from me, was to raise your prices. Low prices attract low-value customers, charge more, target businesses, and so on. I still believe that in general.
But the situation has changed. When the supply of software outstrips demand for it, the price of a feature trends to zero whether you like it or not. In that world there's a real argument for getting there first: lower your prices deliberately, capture as much of the market as you can while competitors are still defending their margins, and make the revenue back with add-ons, usage and upsells.
I'm not saying race to the bottom. I'm saying that if your core product is becoming a commodity, don't price it like it isn't. Own the market, then sell them the things that aren't commodities yet.
7. Become the all-in-one
Why keep paying eight subscriptions when a developer can now build each of those tools in a week? Plenty of companies are already doing exactly that, replacing individual vendors with internal tools one by one.
But there is an inflection point coming. Maintaining eight internal tools is not practical for most companies, even though they still want the upside of building their own: one login, one set of integrations, one set of permissions, one bill. That is the opening for an all-in-one vendor, a single product that covers all eight, so the company gets the consolidation without the maintenance.
This is part of where Bannerbear is going. Image generation, animation, video tools, PDFs and workflows in one API, because the alternative for our customers is stitching together a handful of services, or building them. Being the one thing that replaces a pile of things is a solid moat, as long as you actually do all of it well.
Pick two or three
You can't do all seven. Some contradict each other. But most SaaS companies can do two or three, and I think that's enough.
What I'm sure of is that the old answer, build more features, is not on the list. That era is over, and honestly, I don't miss it. It was never the interesting part anyway.
If you're a founder thinking about this stuff, I'd love to hear which way you're leaning. I'm @yongfook on X.
