Last week Salesforce, Dialpad and Five9 each announced the purchase of an AI vendor. CX Dive reported it on 11 November 2026. Salesforce is buying Cresta, a voice AI agent company. Dialpad is buying Ada, which runs AI agents across voice, email and social. Five9 is buying Level AI, which was already its partner and sold in its marketplace. The article gives no deal prices.

What the excitement says

The reading you will see most is that AI is now so valuable that the big suites have to own it. There is some truth in that. Terra Higginson of Infotech says buying proven capabilities cuts time to value for the buyer.

I test a claim like this by asking what actually changed. The suites wanted AI agents and bought them instead of building them. That tells me the AI chat layer is something a large vendor can acquire in a week. It also tells me a standalone chat layer has a hard time staying standalone.

What did not change

Your customers still ask the same things. Where is my order. Can I change my booking. Can I get a refund. None of that moved because three companies signed deals.

Every AI chatbot is almost the same at the chat level. They read the question and write a decent answer, and they do it in many languages. If “has AI” is the only selling point of a support tool, it is not a selling point. Three large buyers just showed what a business with only that pitch is worth to a bigger company. It is worth being folded into someone else’s product.

What I would judge a support tool on

Take a small clinic. A patient messages at 9pm to move an appointment. A tool that answers “our hours are 9 to 6” has used AI and resolved nothing. A tool that checks the diary, offers three slots, moves the booking and sends the confirmation has resolved the conversation.

Or take an online shop. A customer asks where the parcel is. The useful tool looks up the order in your store and replies with the status. It does not link to a tracking page and call that handled.

So I would ask two questions about any vendor:

  • How many conversations end resolved, with the customer’s problem actually solved? Ask for this in a form you can check with your own data.
  • Can it act in your own systems? That means orders, bookings, payments, delivery and handover to your team with the details attached.

A demo that shows a clever reply answers neither question. Ask to see it do the job on your real workflow.

The contract risk

The CX Dive piece raises the part I think matters most to small operators. Rebecca Wettemann of Valoir asks whether a product you have invested in will keep being supported once the vendor is acquired. She advises planning for that at the start of a contract, not when the announcement lands. Both analysts expect more consolidation, which could leave buyers with less choice. Higginson also warns that if you pick one vendor, you may become increasingly all in.

For a business without a procurement team, that translates into simple habits:

  • Prefer monthly or annual terms over multi-year commitments with any vendor whose main pitch is AI.
  • Check that you can export your data, your conversation history and your configured workflows.
  • Ask what happens to your plan, your pricing and your support if the vendor is bought.
  • Keep your customer data and rules in systems you control, so a switch costs you weeks and not months.

None of this assumes your vendor is next. Cresta, Ada and Level AI all have real customers, and the article does not say they were in trouble. The point is that you cannot know which vendor is next, so the contract should not depend on it.

Where this touches what I build

This is the reason I describe AeroChat as chat infrastructure and not as a chatbot. The chat layer alone is the part that gets copied and bought. What is harder to copy is the work underneath. A conversation has to become an action in your systems, such as checking an order, booking an appointment or sending a payment link. In my view, the local payment and logistics connections matter too, because a global vendor often skips them.

What to do this month

Open your current support tool’s contract and read the term, the exit clause and the data export terms. Then pull a sample of last month’s conversations and count how many ended with the customer’s problem solved, with no person needed. If you cannot get that number from your vendor, treat that as an answer.

When you shop for a replacement or an addition, drop “has AI” from your checklist, since every vendor will say yes. Replace it with a test on your own workflow, run end to end, and sign for the shortest term they will give you.

Source: CX Dive: CX vendor consolidation: A wake-up call for CX leaders