WhatsApp Blue Badge

WOZTELL’s ChatGPT Extension Update: What It Means for Enterprise

by Jamie Chew | May 18, 2026
WOZTELL’s ChatGPT extension just got smarter. For the enterprise AI chatbots Sanuker deploys, here’s what now changes — search, retrieve, and act.

We’ve just rolled out a major update to the ChatGPT extension on WOZTELL, our conversational AI platform. For the enterprise teams Sanuker works with — running customer-facing AI chatbots on WhatsApp, Instagram, Messenger, or Webchat — this update changes what’s possible inside a single conversation.

The updated extension is powered by OpenAI’s Responses API. It gives the AI chatbot three new tools out of the box: searching your knowledge base, pulling real-time information from the web, and connecting to your CRM, ERP, or other core systems mid-conversation. Together, these capabilities move the chatbot from one that replies to one that acts.

If your enterprise is already running an AI chatbot deployment, there’s also a migration deadline to plan around: OpenAI is sunsetting the previous Assistants API on 26 August 2026, after which the old version will stop working.

Here’s what’s new in the extension, and what enterprises should be planning for.

What changed inside the extension

The previous version of the ChatGPT extension on WOZTELL was built on OpenAI’s Assistants API. That version handled prompts, tone, and conversational logic well — but reaching beyond the prompt into your business data required significant custom development.
The updated version, built on the Responses API, comes with three tools the AI chatbot can use natively. For enterprise deployments, Sanuker configures these to your data sources, systems, and governance requirements during onboarding.

The shift is from a chatbot that replies to one that acts.

Three capabilities now available in enterprise deployments

File Search: It can search your own documents before answering

The AI chatbot can now query a connected Knowledge Base — your product catalogues, pricing sheets, policy documents, technical manuals, internal SOPs — and ground its replies in the retrieved content. Instead of stuffing everything into a static prompt, your enterprise knowledge lives in a structured, searchable library the chatbot draws from on demand.
In the enterprise deployments Sanuker builds, this is where most of the upfront consultancy work sits. A knowledge base that performs reliably at scale needs document curation, version control, segmentation by audience and language, and an update process that keeps content fresh without breaking citations. Done well, it eliminates the “let me transfer you to an agent” answer for the majority of inquiries.

Web Search: It can look things up on the web

For industries where public information shifts often — interest rates, regulatory updates, product availability, shipping schedules — the AI chatbot can now pull current information from the web at the moment of the conversation.
We typically restrict these searches to a controlled list of domains: your own support site, official regulatory portals, verified industry sources. This keeps results relevant and avoids the open-web risks enterprise compliance teams rightly worry about.

Function Calling: It can connect to your core systems mid-conversation

This is the capability that turns a conversational chatbot into an operational one. With Function Calling, the AI chatbot can check an order status against your ERP, look up a customer record in your CRM, validate an appointment in your scheduling system, or create a support ticket in your service desk — all without leaving the conversation thread.
Each function is a custom integration with an existing enterprise system, with authentication, rate limiting, error handling, and audit logging built in. It’s also the capability that delivers the highest measurable ROI in most of our deployments — automating the lookups that previously required a human agent to switch tools mid-conversation.

Two deployment modes, one extension

Not every customer interaction should be fully automated. The updated extension supports two modes that can run side by side on the same WOZTELL setup, governed by escalation rules you define.

AI Chatbot for full automation

For high-volume, predictable inquiries — order status, appointment booking, FAQ resolution, basic troubleshooting — the AI Chatbot replies directly to customers on WhatsApp, Instagram, Messenger, or Webchat. No human in the loop unless escalation rules trigger one.

AI Inbox Replies for human-in-the-loop

For complex, regulated, or high-stakes conversations — pricing negotiations, claims discussions, sensitive support cases — AI Inbox Replies generates suggestions inside the agent’s inbox on WOZTELL. The agent reviews, edits, and sends. Speed gains without removing human judgment from the loop.
Both modes draw on the same Knowledge Base, web search, and system integrations. The decision is operational, not technical.

For existing AI chatbot deployments, migration is required by August 2026

If your enterprise is already running an AI chatbot on the previous version of the extension, there’s a deadline to plan around. OpenAI is sunsetting the Assistants API — which powered the previous version — on 26 August 2026. After that date, any chatbot still on the old version will stop working.
Migration is not automatic. A new extension needs to be configured, your existing prompts and logic carried across, and parallel testing run before cutover. The good news: your Instructions, model settings, and core logic can be replicated, and running both versions in parallel during the transition is supported — as long as they’re not connected to the same channel at the same time.
Sanuker handles this migration end-to-end for existing enterprise clients, including knowledge base setup, function rebuilding, and channel-by-channel cutover with rollback in place. Teams with high-volume deployments should aim to be in execution by Q1 2026 to leave room for thorough testing.

Talk to Sanuker about deploying or migrating your AI chatbot

Whether you’re scoping an AI chatbot deployment on WOZTELL for the first time or planning your migration before the August 2026 deadline, Sanuker handles the consultancy, configuration, knowledge base architecture, system integrations, and change management end-to-end.
Book a consultation with our team to map out what a deployment or migration looks like for your business.