Applied AI - key considerations and lessons learned

Algospark has been delivering successful applied AI systems since 2015. We have worked across numerous industry sectors and have learned how to quickly deliver value and avoid common AI project pitfalls. In the following paragraphs we share our learnings over multiple successful project deliveries.

Clear scope

• There is wide scope for using AI (innovation and process efficiency)
• Use clarity in scope and purpose (why, what and how)

Planning

• Set modest early goals initially focused on a subset of products for part of the business
• Involve service “end users” early and iterate deliverables frequently
• Use phasing (early prototype / static example / dynamic solution)

AI has to be business area led - not technology led and needs to be supported by a proficient data and IT team. Don't start in house if you do not have a solid AI team. It is best to use applied AI specialists to work on early ideas and develop an applied roadmap. Remember, early AI is experimentation, so keep proof of concept projects focused and short. Once early prototypes demonstrate value, transition to in-house capabilities as the projects grow and solutions can then be rolled out across the business.

Develop applied AI solutions as a standalone capability - avoid developing an integrated system capability from the outset. In other words, develop AI in a sandbox and where possible, use cloud systems that can easily be set-up and scaled or turned off. Do not let technical jargon or existing data and technology get in the way.

Team

• Good high level design skills are key: what is it, why is it valuable, how can minimum viable product be developed
• Keep teams small, but ensure a strong mix of skills spanning business understanding, AI and application development
• Keep learning! Things change quickly.

Engage experts - it is unlikely you will have the upfront design and implementation skills. Make sure that your delivery team includes skills spanning :

• Business understanding: business process understanding, analysis skills (where are opportunities) and high-level business case (finance) understanding
• Data analysis and engineering: understanding and impact of data stores (SQL/JSON), data pipelines (ETL) and dashboards (PowerBI / Tableau)
• Algorithms, models and applied maths: know the best approaches of model selection, considerations and testing. Knowledge of Python (possibly also R).
• Cloud management: set-up of data storage, compute, networking, administration and security.
• Application development: putting models into use (Python and front-end web skills)
• Project management: agile planning, managing, monitoring and reporting

We hope this helps. Please reach out to us at Algospark if you would like to discuss how you can get the most out of applied AI.