There is understandably a lot of interest in adding AI to existing applications.
For some organisations, that might mean introducing intelligent search. For others, it could be assisted reporting, automated analysis or AI-supported workflows.
And in some cases, adding an AI capability to an existing application can be relatively straightforward.
The more difficult question is what happens when AI needs to work more deeply with the application, its data and the way users already work.
That is where the foundations matter.
If information is difficult to access, systems are difficult to integrate, permissions are unclear or the underlying architecture makes change difficult, introducing AI can become much harder than the initial use case suggests.
So before asking which AI capability to introduce, it is worth asking a more fundamental question:
Is the application ready to support it?
AI Readiness Starts Before the AI Model
AI readiness is not simply about choosing an AI model or connecting an AI service to an existing application.
Some AI capabilities can be added without significant changes to the underlying platform. But deeper integration often depends on foundations such as well-structured data, reliable interfaces, appropriate permissions and architecture capable of connecting information safely.
These foundations determine what an AI capability can access, how it can be integrated into existing workflows and how confidently the organisation can govern its use.
This is particularly relevant for legacy applications.
An application may have accumulated years of business logic, data structures, integrations and workarounds. It may still perform its core function perfectly well, but that doesn’t necessarily mean it is ready for new capabilities.
AI therefore becomes another reason to look at the application as a whole.
Not because every application needs to be modernised before AI can be introduced, but because organisations need to understand where the existing platform is sufficient and where its limitations could constrain what they want AI to do.
The following seven questions provide a useful starting point.
7 Questions to Assess AI Readiness
1. Is your data structured and accessible?
AI capabilities depend on access to useful information.
That sounds obvious, but many business applications contain information in different structures and locations. Some may sit within the application itself, while other information may exist in documents, spreadsheets or separate systems.
The first question is therefore not simply whether you have data.
It is whether the information required by the AI use case can be accessed, understood and used appropriately. That also means understanding the quality and provenance of the information, and confirming that its use is permitted for the proposed AI capability. Accessible data is not necessarily accurate, trustworthy or appropriate to use.
If the answer is no, improving the data foundation may need to happen before or alongside the AI work.
2. Can systems communicate through reliable APIs?
AI rarely operates in complete isolation.
Depending on the use case, it may need to retrieve information from an application, return results to it or interact with other services.
Reliable interfaces make those connections easier to manage.
If an application relies heavily on tightly coupled components, bespoke integrations or interfaces that are difficult to work with, introducing new capabilities can become more complicated.
This doesn’t automatically mean the application needs to be rebuilt.
It does mean the integration architecture should be part of the AI-readiness assessment.
3. Are permissions and identity clearly managed?
AI may need access to information that isn’t appropriate for every user.
That makes identity and permissions particularly important.
An organisation needs to understand who can access what information and how those rules are enforced when new capabilities are introduced.
A useful AI capability should not create a new route around existing access controls.
The question is therefore whether the application’s existing identity and permissions model provides a sound foundation for introducing AI safely.
4. Can the architecture scale?
An AI capability can change the way an application is used.
A search function, for example, may introduce new patterns of interaction or require access to information across different parts of a platform.
The application therefore needs an architecture that can support the expected demand and future evolution.
Scalability doesn’t necessarily mean adopting a particular architecture or technology.
The appropriate architecture depends on the application and the business.
The more useful question is whether the current architecture gives the organisation enough flexibility to support the capabilities it wants to introduce.
5. Can information be surfaced across systems?
One of the most useful applications of AI can be helping people find and make sense of information that is already available but difficult to navigate.
That becomes much harder when information is fragmented across different systems and sources.
Before introducing AI-driven search or analysis, ask:
Where does the information live, and can the application access the relevant information in a consistent way?
If users currently need to search multiple systems, documents or repositories manually, there may be an opportunity for AI to improve that experience.
But the AI capability will still depend on the quality and accessibility of the underlying information.
6. Can AI functionality be tested and governed?
Introducing AI doesn’t remove the need for software engineering discipline.
AI-generated or AI-assisted outputs need to be tested. Access needs to be controlled. The organisation needs to understand how the capability behaves and where human review remains necessary.
Outputs should be grounded in trusted information and traceable back to their sources wherever the use case requires it. The organisation should also define where human review takes place, who remains accountable for the resulting decisions, and how cost, output quality and changes in behaviour will be monitored once the capability is in use.
This is particularly important where AI is supporting business processes rather than simply providing a convenience feature.
The architecture, engineering and business teams therefore need to consider how the AI capability will be tested, monitored and governed before it is introduced.
AI can accelerate parts of development and investigation, but it doesn’t replace human judgement.
7. Can the application evolve without major redevelopment?
Perhaps the most important question is what happens after the first AI capability is delivered.
If adding one capability requires significant redevelopment across the application, the organisation may find that every subsequent change becomes another major project.
A modern application should provide enough flexibility to evolve as business needs change.
That doesn’t mean predicting every future AI use case.
It means avoiding an architecture where every new capability requires the organisation to revisit the foundations from scratch.
This is one of the strongest arguments for considering AI readiness as part of wider application modernisation, rather than treating every AI initiative as an isolated project.
Why Adding an AI Layer Is Not Always Enough
There is a temptation to think of AI as something that can simply be placed on top of an existing application.
Sometimes it can.
A focused AI feature may be introduced without changing much of the underlying platform.
The challenge comes when the organisation wants AI to work deeply with existing data, processes and user journeys.
At that point, the limitations of the underlying application become much more relevant.
Technical debt
Years of workarounds and tightly coupled components can make new development more difficult.
AI doesn’t remove that complexity. In some cases, it can expose it.
Poor data foundations
If the information required by an AI capability is fragmented, inconsistent or difficult to access, the quality of the resulting experience will be constrained by the underlying data.
Integration complexity
An AI capability may need to interact with multiple parts of an application or other systems.
Where interfaces are unreliable or difficult to work with, that integration becomes another source of complexity.
Security and governance
AI needs to operate within appropriate permissions and governance arrangements.
The organisation needs confidence that users are seeing the information they are entitled to access and that AI-assisted outputs are being used appropriately.
The point isn’t that every one of these problems needs to be solved before introducing AI.
The point is to understand the dependencies before committing to the solution.
AI readiness is therefore less about whether you can add AI and more about whether you can integrate it in a way that remains useful, secure and maintainable.
Modernisation Creates the AI Foundation
When deeper AI integration does require changes to the application, it helps to think about the foundations as a connected chain:
Architecture → Data → APIs → Cloud → Governance → AI
Architecture determines how the application can evolve.
Data determines what information AI can work with.
APIs and other interfaces determine how systems and capabilities can communicate.
Cloud or other appropriate infrastructure provides the environment in which the application operates and evolves.
Governance establishes the controls around security, access and responsible use.
AI then becomes a capability that can be introduced where it creates genuine value.
This is why AI application modernisation should not be reduced to an exercise in adding AI functionality.
The objective is to create an application that can support the capabilities the business needs now while remaining flexible enough to evolve as those needs change.
And importantly, modernisation doesn’t necessarily mean adopting one particular technology architecture. The right approach depends on the application, its requirements and the organisation’s future plans.
Twinnedit: Building the Foundations for AI
Twinnedit provides a useful example of this approach.
The building assurance and safety-management platform was functional, but its existing technology, user experience and data structures were increasingly limiting what the business could create next.
The decision was made to rebuild the platform rather than continue making incremental improvements to the existing foundations.
Neem worked with Twinnedit’s internal team over approximately six months, beginning with new UI and UX designs and focusing on user journeys rather than simply reproducing the existing application.
The rebuilt platform introduced a responsive Vue 3 front end, a .NET Core API and an Azure-based architecture using Azure App Services, Azure SQL and Azure Blob Storage.
The significance of that work for AI was not simply that the application had a newer technology stack.
The rebuilt architecture provided a stronger foundation for future development and enabled practical AI capabilities.
One example is global AI-driven search, allowing users to find information across multiple parts of the application through a single experience. The platform also supports assisted safety case reporting, bringing together relevant evidence and organising it against known reporting requirements.
This is a useful illustration of the distinction between adding an AI feature and creating the foundations for deeper AI capability.
The objective wasn’t to introduce AI for its own sake.
The platform was modernised because its existing foundations were becoming a constraint, and the resulting architecture created opportunities to introduce capabilities that were more difficult to support previously.
Twinnedit demonstrates that AI readiness can be an outcome of broader application modernisation rather than a separate technology exercise.
AI Readiness Checklist
Before adding AI to a legacy application, ask:
Data
- [ ] Is the data required for the AI use case structured?
- [ ] Is the quality and provenance of the data understood?
- [ ] Is the data permitted to be used for this AI capability?
- [ ] Can the relevant information be accessed?
- [ ] Do you understand where the information currently resides?
Architecture
- [ ] Can the application support the proposed capability?
- [ ] Can it evolve without significant redevelopment?
- [ ] Are existing architectural constraints understood?
Integration
- [ ] Can relevant systems communicate through reliable interfaces?
- [ ] Can the AI capability access the information it needs?
- [ ] Are existing integrations sufficiently understood?
Identity & Permissions
- [ ] Are user identities clearly managed?
- [ ] Are access permissions well defined?
- [ ] Will the AI capability respect existing access controls?
Governance and Operations
- [ ] Can outputs be grounded in and traced back to trusted sources?
- [ ] Is human review defined, with clear accountability for decisions?
- [ ] Can cost, output quality and behavioural changes be monitored?
Scalability
- [ ] Can the application support the expected usage?
- [ ] Can the architecture accommodate future development?
- [ ] Have potential performance implications been considered?
Governance
- [ ] Can the AI functionality be tested?
- [ ] Can its outputs be reviewed where necessary?
- [ ] Are security and governance requirements understood?
- [ ] Is human oversight required for the intended use case?
Future Evolution
- [ ] Can additional AI capabilities be introduced later?
- [ ] Does the proposed solution avoid creating another tightly coupled dependency?
- [ ] Is the application becoming more adaptable rather than simply adding another feature?
If several answers are “no”, that doesn’t necessarily mean the organisation should stop its AI initiative.
It may simply mean that the AI project needs to include some application modernisation work alongside the capability itself.
AI Readiness Is a Business Question, Not Just a Technical One
The question “Is our application AI-ready?” can sound like a technical assessment.
It is partly that.
But it is also a business question.
What does the organisation want AI to achieve? What information will the capability need? How will users interact with it? What needs to happen behind the scenes? What controls need to be in place? And how easily will the application be able to evolve after the first capability has been introduced?
Some applications will be ready for AI without significant modernisation.
Others may need targeted changes to data, interfaces or specific components.
And for some, the limitations will be broad enough that a wider modernisation programme makes more sense.
The important thing is to understand which situation you’re in before deciding what to build.
Twinnedit is an example of a platform where rebuilding created the foundations for a more capable product, including AI-driven search and assisted reporting. The lesson isn’t that every legacy application needs to follow the same path.
It’s that AI readiness should be assessed against the application, the business problem and the capabilities you want to introduce.
Is your application ready for AI?
Neem’s Application Modernisation Discovery Session can help you understand where your existing application is creating constraints, what can be modernised incrementally and where a broader transformation may be appropriate.
Assess whether your application is ready for AI.









