Our AI Metrics & Process Improvement services help you create the organisational capability to build AI/ML solutions reliably and cost-effectively. This may be using in-house teams, outsourced development specialists, or a mixture. We are also agnostic about the use of AI/ML packages and platforms vs. bespoke development of your AI/ML solutions.
Our AI Metrics & Process Improvement services cover three main areas:
Choosing an appropriate AI/ML capability approach for your circumstances
Implementing effective, scaleable development processes for AI/ML work
Defining pragmatic metrics to manage & monitor AI/ML work without impeding it
Building AI/ML solutions is following a similar evolution to software development:
Initially a “dark art”, software was originally hand-crafted by skilled (expensive) programmers
It then became an engineering discipline, based on processes & methodologies. This made it more reliable, with both skills and suppliers becoming widely available.
The evolution culminated (for now) with mainstream adoption of tools, services & packages. This relegated much software (& related skills) to commodity status, especially with the availability of end-user tools.
AI and ML could be considered still to be in its equivalent of the first stage. But while that’s probably true for most businesses, it’s already possible to move an individual organisation to the second stage. And not necessarily that difficult.
We help you move to a process-based approach to AI/ML work, allowing you to incorporate new tools, services and packages. We do this through capability, processes and metrics.
Choosing Your Approach to AI/ML Capability
Three key decisions to be made are:
Will AI/ML be core to your business, or is it just a technology to exploit?
Package/platform vs bespoke development? (Short & long term)
In-house vs outsourced vs augmenting your team with external specialists?
If you already have reasonably robust IT development processes and tools, these will form a good basis for your AI/ML work.
Some AI/ML work is relatively standard software development; your current processes will probably be fine for those with relatively little change. Other work is very different in nature, such as algorithm development, so additional processes may be required.
However, one of the trickiest areas is AI/ML work that is superficially similar to modern software development, but in practice quite distinct. For example, many organisations have adopted Agile for their software work, but they way they apply it wouldn’t work well for AI/ML.
There are two types of metric to consider for AI/ML work:
Development metrics to monitor project progress and assure product quality
Implementation metrics to ensure the AI/ML solution is working as intended, and ideally to demonstrate the business impact and value
Some organisations – particularly those with sophisticated IT departments – will track software development metrics such as defects and productivity. Others may only work at a higher level, focusing on product backlogs and feature burn-down rates.
The role of our AI/ML Metrics work is to find the appropriate level of measurement & monitoring for what an organisation is familiar with. As with other types of metric, the human factor is key to whether these are successfully introduced, and whether they end up helping or hindering.