A small-looking product update is signalling a big behavioural shift: people will increasingly discover, evaluate, and transact inside AI agent interfaces. Both externally with customers, and internally with employees.
If you still think “digital strategy” starts and ends with optimising your website and a mobile app, it’s time to update the mental model. The new front door is becoming a conversation.
What Just Happened (And Why It Matters)
OpenAI confirmed that developers can submit apps for review and publishing inside ChatGPT, and that an in-product App Directory now exists where users can browse and discover them. OpenAI also released an Apps SDK (beta/preview) to build chat-native experiences.
In business terms, “ChatGPT Apps” introduce a new distribution and interaction layer:
- Users don’t navigate to pages and menus; they ask, and the app does.
- Workflows become chat-native: guided conversations that can fetch information, take actions, and complete tasks.
- Publishing starts to resemble a “store” model. Metadata, quality, safety, and review expectations become part of how you win attention.
Early coverage suggests rollout will be gradual into the new year, and developers can manage where apps are available (useful if you’re thinking about UK/EU vs Asia vs US rollouts). There is already clear scale to make businesses think differently - OpenAI’s latest numbers released at DevDay indicated ~800m weekly active ChatGPT users.
Your “Findability” Moves From Pages To Answers
For two decades, the assumption was simple: get people to your site/app, then help them find what they need. In an agent-first world, discovery becomes: be the best answer and the fastest action inside the agent. That changes how demand is captured and how businesses decide to take products to market:
- From SEO to “agent optimisation”: the agent decides what to recommend, what to compare, and which action to take.
- From navigation to completion: fewer clicks, fewer drop-offs, less “browse fatigue.”
- From funnels to conversations: intent is expressed in plain language, in context, often closer to purchase.
It’s not that websites disappear. It’s that they become less of the primary interface, and more of a backend, reference point, or fulfilment layer.
Why This Matters To Customer Experience
When the AI-based interface can complete the transaction, a lot of the normal friction for the client disappears.
Early adopter retail brands are already treating this as a channel, not a novelty. In the US, Target has announced a curated shopping experience inside ChatGPT, including browsing, multi-item baskets, and fulfilment options, positioned as a new way to shop. And Walmart has announced a partnership with OpenAI to enable AI-first shopping via ChatGPT.
This is the signal: major brands are planning for ChatGPT as a customer channel with its own merchandising, conversion, and measurement playbook. Here are some examples we can see playing out:
A Quick Retail Scenario
A customer asks in ChatGPT: “I need everything for a 10-person vegetarian dinner party on Saturday so give me quick recipes, wine pairings, and delivery slots after 6pm.” A chat-native shopping app can propose menus and substitutions, build a basket in one flow, offer fulfilment options, and complete checkout, without the customer ever “entering” a traditional site experience. The same flow will work for insurance, travel, clothes, etc.
The customer no longer navigates multiple websites but instead has an outcome with the least friction possible.
A b2b Scenario: Your Digital Presence Must Be Actionable Inside An Agent Layer
An independent financial adviser is in a client meeting and uses ChatGPT connected to a fund platform. The IFA will do its AI-powered scenario planning with and for their client to have plain language interactions before simply saying “So now switch from Balanced to Defensive, increase cash buffer to 18 months of withdrawals, and rebalance across their ISA and SIPP. Draft the suitability note and prepare the order for approval.”
At that point the app can translate the conversation into platform actions: propose the trades, check mandate constraints and suitability rules, surface any compliance flags, and route the final step for human sign-off. Once approved, it submits the orders straight into the fund platform.
The key shift is that advice, documentation, and execution happen in one continuous flow, rather than bouncing between modelling tools, PDFs, email, and separate dealing screens. Your digital presence becomes how well your product can be evaluated and actioned inside an agent layer.
Move Workers From “Time To Answer” To “Time To Action Completed”
Inside organisations, ChatGPT Apps can become a guided “do my job” layer that sits on top of company specific policies, knowledge bases, and systems, while maintaining governance.
High-value internal journeys are predictable:
- HR / IT workflows: onboarding, access requests, device setup, benefits
- Case handling: triage, drafting responses, pulling prior cases, suggesting next steps
- Sales enablement: account briefs, proposals, RFP Q&A, battlecards
- Policies / SOPs: “What’s the approved process for X in the EU vs APAC?”
The productivity gain is shifting from time-to-answer to time-to-action: fewer handoffs, fewer “ok, now I know which system to use” loops, and more seamless execution of standard processes.
Because Apps are built with defined behaviours and connections, you can design for access controls, auditability, and clear guardrails on what the app can/can’t do.
Why Moving Early Actually Matters
Getting in early is about building real advantages while the space is still forming.
- You become the default choice: When users are exploring a new directory, the apps they find first tend to stick. More than just another option, you're the one they try first and build their habits around.
- You learn what people actually need: You'll see the real questions customers ask, how they phrase things, what trips them up. That means you can build around actual behaviour instead of guessing.
- You create habits that are hard to break: Once your app becomes the fastest way to get something done, whether that's making a purchase, booking something, or solving a problem, people stop looking for alternatives.
- You spend less on customer acquisition: When the agent handles the heavy lifting, you're not paying as much to guide people through complex processes that used to need human support.
- You build something competitors can't easily copy: The real edge isn't just answering questions better. It's having your specific data, your refined workflows, and the feedback loop that keeps making everything sharper.
The risks for laggards are equally concrete:
- Disintermediation: the agent becomes the interface; your site becomes a backend utility.
- Losing the narrative: if third parties explain your offering better inside the agent, you lose positioning control.
- Commoditised comparison and higher acquisition costs: if you’re not present in the agent channel, you pay more to pull users back to legacy journeys.
What To Do In The Next 30–90 Days
The pragmatic path is to start with two journeys, one external and one internal to build the skills and confidence.
- Pick 1- 2 high value journeys: For example, a customer journey from discovery to basket/quote to checkout/lead capture and an employee journey such as onboarding, policy-to-action workflows, or service desk handling.
- Define success metrics upfront: Adoption, conversion or lead quality, resolution time/deflection rate, time-to-complete.
- Confirm data readiness and ownership: What sources are required (catalogue, pricing rules, SOPs, knowledge base)? Who owns accuracy and updates? What gets excluded from early pilots?
- Set guardrails: What can run autonomously vs what needs approval, escalation paths to humans, logging and monitoring expectations.
- Pilot > measure > iterate > scale: Ship quickly and treat the first release as an instrumented learning loop, not a big-bang migration.
How Tomoro Can Help
Tomoro is an AI strategic and engineering consultancy. We have deep knowledge and experience, build on our in-house R&D function which partners closely with OpenAI R&D. We design and build “first-in-the-world” AI solutions as part of our clients' AI transformation journey. ChatGPT Apps is part of that, with agent workflows for internal productivity and customer experiences. Here’s how a typical “idea to live” cadence works in 8-12 weeks:
- Week 1: use-case framing, success metrics, data/knowledge audit
- Week 2: prototype experience, content shaping, interaction design
- Week 3-9: build, integration, testing, governance checks
- Week 9-12: pilot launch, measurement plan, iteration backlog
If you're interested, please get in touch.
So if you’re building your 2026 technology roadmap, the key question is no longer about optimising your website or app, but “How well can customers and employees get outcomes through an agent interface, safely, fast, and at scale?”
