Key takeaways
Most AI projects get watered down to Yet Another Answering Service
Conversation Ops maps how work actually happens, not just chat
Connects systems of record so conversations turn into action
Builds toward autonomy in stages, with human oversight throughout
We started AllSet.solutions to give organisations the confidence to autonomously run high-value conversations, ones that benefit all parties, and where learnings continuously improve confidence and quality.
This sounds easy, but we know it's a big ask.
We all know how hard it is to get a product or service into production within a large, complex organisation. AI is a strategic pillar, so some kind of work has to happen. But everyone has an opinion and, between the AI slop, failed POCs and never-ending game of AI product whack-a-mole, those opinions are somewhere between scared and sceptical. Ideas are eroded in the meeting room, problems are found and the final agreement is the absolute minimum everyone can live with. So that gets done.
And that's how you got Yet Another AI Answering Service™.
Meanwhile, the business carries on running through conversations. People agree things in chat, chase missing information and explain what they need. Much of that never reaches the systems the organisation relies on to understand its own operations.
Conversation Ops starts by getting underneath this: understanding how the work actually happens and putting the data in place to support it. That gives the organisation a basis for managing those conversations with conversational AI, automating unnecessary work and having useful conversations it couldn't reliably have before.
Getting underneath the operation
We need to follow the work as people actually do it. Where does someone have to ask a colleague for an answer? What gets agreed in a chat and then copied into a system? Where does the process stop until a particular person is available?
This takes data engineering as well as time with the people doing the job. We connect the relevant systems of record to bring together facts such as bookings, qualifications and availability. Alongside that, conversations need to be monitored, properly analysed and stored so we can capture what was learned. Those learnings might explain a decision, record a preference or tell us why the information in a system no longer reflects what's happening.
Both need to be reliable and traceable, with access and retention governed by the organisation's rules. We need to be able to check which facts and learnings informed a decision.
Put those facts and learnings together and you can understand enough of the operation to act. They inform how a conversation is handled, what can be done as a result and when someone needs to step in. What happens in that conversation then feeds back into the next decision, whether it involves a customer or a colleague.
Seeing where the work gets stuck
Once that's in place, you can see where teams spend their time chasing answers or fixing the same problem repeatedly. Conversations start to show you where the operation is failing and what it costs people to keep it going.
You can then decide what to do about it. A conversation that needs judgement can be supported with the right information and a clear route to a decision. If it only exists because someone has to move information between systems, that work can be automated.
With staffing and workplace coordination, for example, qualifications and availability can inform who is approached for a shift. Their response can lead to an agreed booking and an updated rota, with a manager involved where needed. You can monitor whether cover was arranged and where it got stuck.
You can also see why the same staffing problem keeps coming back. There's a chance to address the cause, instead of getting very efficient at sorting out Saturday, every Saturday.
Article continues below
What is Conversation Ops?
Conversation Ops is the operational layer underneath an organisation's conversations. It maps every channel a prospect or customer might use, monitors what happens inside them, connects that to the systems of record that already run the business, and keeps people in the loop at every stage of automation.
Having the conversations you couldn't have before
For customers and members, the same foundation lets an organisation use what it knows across different systems and previous engagements. It can identify someone who would benefit from a conversation and have enough context to carry it through properly.
In premium retail, that can mean building a taste profile from purchase history and conversations with advisors. When a particular piece arrives, you can find the customers likely to value it and approach them with something relevant. We've demonstrated this capability in clienteling.
In a fitness club, it means connecting a trainer's new qualification with members whose goals make them a good match, then making an introduction that works for both. Across a large membership, that's difficult to do when the qualification sits in one system and what the member wants is buried in a conversation somewhere else.
As people respond, the organisation learns more. That information should shape what happens next, including knowing when to leave someone alone. The next conversation can pick up from what the organisation has already learned.
Having the confidence to act
Since early 2026, we've been asking organisations: 'Would you let an agent find a customer and have a chat with them?'
The response is usually a haunted-owl-face look.
That's because letting an agent loose on customer data is scary enough. An unscripted conversation is terrifying. What if it says the wrong thing?
Of course, client-facing staff can say the wrong thing too. We trust them with the work because there are established checks and balances. Someone owns the process, there are limits to their authority, and there's a way to deal with things going wrong.
Conversation Ops needs to make that just as understandable for agents for business. We should know why a conversation happened, what the agent was allowed to do and what actually happened. Someone needs to be able to step in. The data access and security requirements still apply.
We build towards autonomy in stages: dry runs, then human approval, then taking people out of individual decisions once the process has proved itself. That's how it becomes business as usual. The organisation has enough visibility and control to trust it with more of the work.
Why we're doing this
Our founding team has spent decades in enterprise AI, data and operations. We've built operational frameworks, governance systems and large-scale data systems for some of the world's biggest organisations. We understand what it takes to get something into production and have a business depend on it.
Alongside that, we've spent the last two years building and running AllSet.chat, handling bookings, payments and client conversations for service providers. It's still helping people every day. We've had to make conversational AI work with real customers and earn the confidence of the people whose businesses rely on it.
AllSet.solutions brings that experience together. We help organisations put Conversation Ops into practice: work out where the value is, connect the data, build the processes and technology and either run it or hand it over to their teams. Governance and security are part of that work from the start.
If you could see where conversations are keeping your organisation running, and use what it already knows to have the conversations it's missing, what would you do first?
That's the conversation we'd like to have.
Want to see how Conversation Ops works for your business?
Continue reading
More from AllSet
AI ASSISTED SOLUTIONS
What field service dispatch software doesn't fix
A 2024 study of 1,000 companies found most miss the response window. Field service dispatch software doesn't close that gap.
Insights
A paradigm shift in member relationship with Conversational AI
For decades, member data has sat as static facts in a database. Conversational AI turns that record into an ongoing, two-way conversation, running continuously across thousands of members at once
Insights
Three problems facing academies and awarding bodies
Academies, awarding bodies, registers and professional networks share three problems nobody names together: inquiry speed, no real position on AI and no visibility once someone qualifies.