Helping Organisations With Their AI Product Development
Although we see AI in many consumer applications, it has not gained a substantial foothold in a lot of organisations. While AI has the potential to bring real business benefits, there are few products available and a lot of these are not sufficiently mature. How can organisations grow knowledge and confidence in AI implementation? Cyber Smart Consulting’s Lead Data and AIS Consultant, Shirley O’Sullivan explains:
Many organisations are using AI to automate some of their outward-facing operations such as first line customer service. Other players are taking this a stage further, not merely replicating existing processes but using AI to deliver new services and products. They are finding that AI can facilitate genuine improvements in productivity, raising them above the competition. In addition, existing staff, far from being replaced by AI, can take on more productive and fulfilling roles. It is not however all plain sailing.
While there is an emerging wave of AI products, their scope is often limited to industry-standard operations. By contrast, new and niche services demand products that focus on specific sectors, tailored to the needs of an individual business. While there is an increasing demand for AI-related products, the supply side is somewhat limited. Organisations are faced with either adopting limited technology solutions or developing a more bespoke or tailored solution.
AI Product Development is our domain. We help organisations to make sense of the AI solution landscape. Our advisory service guides an aspiring organisation to find the right solution for their business.
What Is Our AI Product Development Advisory Service?
Product management is a key aspect in the development of all today’s successful products, from physical goods to services and software development, especially in the AI sphere. The Product Manager role is critical in liaising with customers, partners, and stakeholders, defining and leading the vision and working with the technology specialists.
We designed our service to help the Product Manager and the development team to execute a successful AI implementation project. The service guides them through a series of clearly defined stages to develop a prioritised action plan; a plan that will deliver the highest value for their unique business vision.
Stages
The stages in the process are:
- Conducting an assessment of technologies
- Developing a product vision and business case
- Defining metrics to verify the product against its objectives
- Scoping the shape, quantity and quality of the data
- Creating a plan for the development of the model
- Developing a plan to monitor the system after deployment
Technologies
AI Product Development involves harnessing specific technologies to deliver better products. The technologies are Machine Learning (ML), Deep Learning (DL) and Artificial Intelligence (AI), often collectively referred to as AI, or more correctly as Autonomous and Intelligent Systems (A/IS). Applying the technologies requires particular skillsets, and it is here that we bring our expertise. As part of an engagement we
- Analyse the use cases that are fundamental to the business and identify those which will return the highest benefit from AI investment.
- Quantify objectives for Return on Investment, success indicators and metrics.
- Scope AI product features.
- Analyse representative data for quality, quantity and duplication.
- Design prototypes and data models using industry best practice.
- Help build Machine Learning tools to analyse patterns and predict events.
- Propose alternative models and compare their likely performance by prototyping and iterating.
What Are The Features Of Our AI Product Development Advisory Service?
Business Case – Following analysis and selection of the most appropriate use cases, it is necessary to focus on creating a very clear and precise business case. This must provide a compelling justification for the project, including the cost, benefit, and risks.
Data Acquisition and Analysis – A key aspect of AI planning is the acquisition and analysis of existing data. The Product Team will need access to sufficient data to capture the range of situations that will be encountered by the use cases identified in the AI roadmap. In many cases, this will need to be sourced from multiple systems and departments across the business, and/or from externally sources. Analysis involves understanding terminology, for example, checking apparently identical entities, duplication, quality, and quantity. In particular, the data must be assessed for size and relevance to the use cases.
Ground Truth Dataset – In the Machine Learning sphere, “Ground Truth” is concerned with the accuracy of the available data. Data will be used to train the model so it must be objective and its accuracy must be proven. Algorithms can then be developed to generate machine learning models and the models can, in turn, be assessed against metrics.
Strategies to overcome bias – Human bias can creep into AI systems inadvertently, for example through training data and assumptions. It is vital to implement strategies to identify and mitigate bias to prevent it from being baked into the models.
Compliance – All industries which process and store data relating to individuals should be well aware of the importance of keeping the data secure and the penalties for failing to do so. The development process must follow a clear strategy for compliance, not only for the final AI system but the protection of sensitive data during acquisition, analysis, and training.
Post-deployment metrics – The AI system must be monitored during live operation by defining objectives and measuring its success against them. Improvements can be trialed using techniques such as “A/B testing”, a process which compares two slightly different variants of the model. Deployment must include a version control process so that any changes can be controlled and documented.
Continuous improvement – Once a model has been approved it can continue to evolve using a process known as Active Learning. Here, the algorithm can interactively query a source of information to generate additional labels. This technique is particularly valuable where there is a shortage of labelled data.
Scalability – The design will necessarily have an initial focus on a small, well-defined functional scope. The business case and product planning must take account of how the system can subsequently be scaled depending on feedback from stakeholders.
What Are The Benefits Of Choosing Our AI Product Development Advisory Service?
Governance – We have the experience and knowledge to ensure there is effective governance of the project, from inception through to operation. This is essential for compliance with external regulations such as data security, but also for transparency and accountability to executive management.
Bias – As an external specialist, we have an objective viewpoint on data and use cases. We can identify where unconscious bias may be present, highlight it and ensure that appropriate strategies are in place to mitigate it.
Customer Trust – Implementing an AI solution must be carried out openly, maintaining and building on the trust that customers have in the organisation and its products. We have extensive knowledge of the legal and regulatory requirements for AI systems. Ensuring that the system is compliant with both external regulations and internal requirements means that the system will have an excellent pedigree, deserving of customer trust. In fact, implementing a successful AI product is likely to gain plaudits from customers.
Methodology – The key to all successful project and product management is the use of an established, road-tested methodology. By adopting our methodology, the organisation will build a repertoire of product, technical and data standards. These will establish a foundation that can be used in enhancing and extending the use of AI in future projects.
Internal Team – By learning on the project, alongside our specialists, existing product and technical staff will play a valuable part in developing the AI solution. This will enhance their value to the business and ensure that the product will continue to be managed and tested by internal staff.
Ongoing Measurement – Ingraining critical success factors and metrics during the development process will ensure they become second nature in the operation phase. Management will continue to have a clear measure of the system’s progress and success.
AI capabilities – The organisation will start to gain the benefits of AI adoption, such as improved productivity and operational efficiency along with faster decision making and better customer experience. This is only the start of a journey, however. A successful project will demonstrate to executives, stakeholders, staff, and customers that AI is a technology that can bring tangible benefits.
With an established familiar methodology and knowledgeable internal staff, extending the application of AI throughout the business will surely follow.
Conclusion
Although we can see AI in many consumer applications, it has not gained a substantial foothold in a lot of organisations. While AI has the potential to bring real business benefits, there are few products available and even these are not sufficiently mature. Further, any commercial AI offering will be generalist in nature, whereas gaining a competitive edge requires an AI system that is unique to the organisation. How else can it enhance the market proposition? An aspiring organisation must have the skills and expertise to cross this chasm.
Using our AI Product Development Advisory Service can assist you on this journey. We can help your organisation to gain a foothold in AI, develop systems that will bring genuine benefit, and grow knowledge and confidence in AI implementation.