Automating Intelligently vs Intelligence *in* Automation

Last week I bagged a free pass to attend TechEx Global, a large IT conference at London Olympia, and promised a write-up. IT conferences have a bit of the air of what I imagine markets in Victorian London, imperial Rome, or late antique Alexandria might have been like – a motley collection of hawkers, preachers, tradespeople, magicians, exotic wares and charlatans, all vying for one’s attention using a range of rhetorical techniques, inducements, demonstrations and legerdemain. I’m happy to report TecheEx was a bustling and interesting affair, despite various TFL meltdowns making getting there and back a challenge!

There were four stages, or themes, with an associated series of speakers, which one could follow. As effective integration and automation are a major part of what Estafet does, I opted for the ‘Intelligent Automation’ stream, nipping to some of the others here and there over the two days. I wondered to what extent, if at all, the hype around AI, and particularly LLMs and generative AI, was impacting the reality of ‘traditional’ systems integration and process automation? This stream appeared to promise an answer to that question.

What rapidly became clear from the get go is that there was a lot of misleading ambiguity in the expression, and ‘Intelligent Automation’ could and should be split into two separate issues – the first could be described as ‘Automating Intelligently’ – i.e. all the well-known challenges faced by organisations seeking to deliver effective process automation (selection, automation, integration etc.), and the second as what I’ll call ‘Intelligence in Automation’ – i.e. the application of AI (in its various forms) within automation. Perhaps not surprisingly, there was a lot of the former discussed at the conference, but not very much of the latter – and this for very good reasons: i.e. the limitations and fundamental misunderstandings around current capabilities of AI as applied to automation.

I’m therefore going to split my write-up into two separate pieces – the first and current one reviewing contributions around the traditional challenges of ‘Automating Intelligently’ and, the second, later this week, to share some of the key (and important!) messages around the nature, limitations and appropriate use cases for AI, particularly LLMs and generative AI.

Speakers addressing ‘Automating Intelligently’ included representatives from Skoda, Asahi/Pilsner Urquell, Ofgem and the Home Office. Their presentations, round tables and Q&As rapidly demonstrated that, away from the near panic and hype around AI, traditional process improvement and automation challenges are still omnipresent despite decades of improved automation tooling and increasingly powerful integration and process orchestration platforms. Topics covered included the appropriate and effective deployment of Robotic Process Automation (RPA), and how to effectively orchestrate automated tasks and processes using various forms of Business Process Management (BPM) and the various integration platforms and technologies available. Within this space, organisations continue to face well-known challenges:

  • How to upgrade, retire or integrate legacy systems?
  • How to select appropriate task and process candidates to automate? With all the traditional answers (rules based, stable, repetitive etc.)?
  • Not simply doing as-is automation but delivering end-to-end process and outcome improvement.
  • Focusing on value chains rather than tasks or even individual processes – i.e. don’t just respond to the request for a bot. Look at the bigger picture.
  • The critical importance of getting business and IT working together, ensuring initiatives are business-led but IT-informed.
  • Ensuring you implement the proper architecture to avoid expensive cul-de-sacs.
  • Proper planning, change management, training and lifecycle management.
  • Whether to use citizen developers or have a centre of excellence (CoE); Some found citizen developers failed due to lack of ownership (Ofgem) while others reported them being very effective if supported by IT (SEG Automotive and Asahi); Yet others much preferred an automation CoE in the form of a central automation team to ensure cohesiveness (Home Office).

What (should not have) surprised me is how little these core challenges have changed and how having an experienced systems integration partner (such as Estafet) with experience of BPM, automation, the building of APIs and effective, measured and appropriate implementation of the range of Cloud and integration platforms and technologies remains critical for organisations seeking to muddle their way through an increasingly complex marketplace of tools, platforms and technologies.

I particularly enjoyed the refreshingly frank and lively talk by Ondrej Cesak, Senior Business Analyst, and Martin Letáček, Head of the Automation Competency Centre at Škoda Auto who, starting only as recently as 2019 with the introduction of RPA and gradually expanding to other capabilities, have been key to delivering a traditional but successful automation / integration journey in their organisation. They pointed out all the usual good practices in this sort of journey already mentioned above: synergy with business, automation of processes rather than provision of bots, importance of business analysis before embarking on a tech. solution, vision, a sound and unified solution architecture, and effective KPIs to demonstrate RoI. They were also honest about where they went wrong: including a messy start with departments going off on their own, creating inconsistent solutions that had to then be resolved, licence (and hence cost) inflation, and the counting of automated processes and bots as a measure of progress rather than initially focusing on improved outputs. In this regard they ended up introducing a useful KPI: # of saved hours per process – which might be useful to others facing similar challenges.

In terms of integration technologies, I noted a growth in powerful IPAAS (Integration Platform as a Service) and automation platforms such as Workato, Nintex and Thinkautomation – unified platforms and automation solutions providing capabilities such as out-of-the-box libraries of pre-built connectors and automation and orchestration tooling – aiming to provide low code / no code platforms. These generate an ongoing challenge to larger traditional Systems Integrators and Cloud providers such as AWS and Microsoft seeking to provide all necessary tools within their own ecosystems but, alongside the advantages they bring, can similarly result in vendor lock-in, so you have to choose the horse you’ll be betting on very carefully.

Another, and important thing I noted, given my original question, was that, despite significant hype and use of AI buzzwords, all of these were using generative AI primarily in the form of code copilots, training assistants and chatbots rather than introducing any revolutionary changes to the work of automation and integration itself.

In the examples I saw, therefore, there was very little Intelligence IN automation, other than the above (e.g. MS Copilot at Asahi) and the use of smart monitoring and devices, e.g. fridges that monitor stock levels (IoT and smart devices are still therefore extremely important in manufacturing contexts). The speakers from Skoda even refused to engage with a question regarding their approach to “AI” – they didn’t want to bring AI into the discussion around intelligent automation. In Cesak’s words: “AI is a hot topic, but do you really need AI? Don’t put the solution first.” Other speakers from the technological branches of the enterprise echoed similarly sober sentiments.

So, to conclude this first piece, the type of ‘intelligence’ needed in automation still remains very much as it has been: organisations continue to need process improvement, technological expertise, and effective change management to solve automation challenges. While “AI” in this undefined manner is a hot topic, when management requests “it” – ensure they understand whether it is needed, in what form, and whether generative AI is the right solution given its limitations.

I will be saying a lot more about the nature and current capabilities of LLMs and generative AI based on some outstanding presentations at the conference in my next piece. These were personally fascinating but also convinced me that a lot of individuals and organisations are probably using these in their current form in completely inappropriate ways, and of the importance of applying their current capabilities to the correct use cases where they can really shine (and avoid the significant problems they can generate when used incorrectly!). I’d like to wrap up with an interesting question in relation to ‘intelligent automation’ posed by James Friend, Director of Digital Strategy, NHS England: “are you making it easy to do the right thing, or hard to do the wrong thing?”

By Paulo Goncalves, COO at Estafet

On a practical note – if you want an experienced, reliable integration and automation partner, give us at Estafet a shout – this is our bread and butter and if you want some non-breathy, unhyped, good, practical guidance for your architects, devs, testers and devops engineers on how to best use ChatGPT within their work, check out the best practice guides produced by our consultants for each of these areas.

Stay Informed with Our Newsletter!

Get the latest news, exclusive articles, and updates delivered to your inbox.