How will the next phase of AI make your enterprise a leader? Automating industrial tasks provides an excellent guide

We are at an exciting tipping point where the hype surrounding artificial intelligence (AI) is now overshadowed by the technology becoming central to diverse applications, and with it comes a new wave of sophisticated automation. Though consumer tasks like cashing a cheque with a smartphone or having an algorithm choose your next movie to watch are familiar applications of AI, the current innovation wave has its eyes set on automating a slew of industrial tasks. As the CEO of Zetane, a Montreal-based company that produces AI solutions for diverse industries ranging from the resource sector, to transport, aerospace and defense, we see firsthand how AI technologies like computer vision solve real-world problems in industry.

The embrace of AI automation, however, shows significant gaps. Most surveys on AI adoption report that even in industries that were first-movers in AI, such as the automotive and assembly and financial services sectors, adoption was generally slow. Falling behind are aerospace, smart cities, environment, energy, and construction. Our conversations with industry leaders mirror these findings, especially amongst business decision-makers in sectors where the use of AI is being considered in high-risk environments where the consequence of failure or bias is grave. What we have seen and what is expected from business decision-makers, government authorities and the general public is that any AI that is released into our lives must be safety tested extensively and proven to do what it claims to do. The ability to easily explain the technology to non-AI experts is also essential. Addressing these concerns provides an accelerant to rapid adoption of more advanced and more impactful AI solutions.

Routine and well-documented

Identifying the best opportunities for AI automation need not be complex. Conduct a critical review of your industry by first identifying repetitive tasks that your business documents well. By “well-documented”, we mean tasks that are tracked, measured and reported using consistent protocols; this documentation–in formats that include texts, location tracking, spreadsheets or images–provides the data needed to train AI solutions. Tally up the amount of time and resources that repetitive task requires. If it is over $ 500 000 annually, the time is ripe to start a proof-of-concept (POC) project with an AI solutions provider. Indeed, the prices for AI solutions are increasingly competitive and now often fall well below the current costs of the task you identified for automation. Our clients have had lots of successes following a POC approach using a small amount of high-quality data. Such preliminary projects serve to validate that a proposed AI solution will bring the benefits promised. Moreover, it helps scope a larger project so that understanding of the technology and data risks prior to launching a larger scale deployment are known.

Industry leads before consumer products: the case of autonomous vehicles

Occupying much media attention today are autonomous vehicles destined for use by the average consumer. Having autonomous vehicles operate in typical urban environments proves to be challenging and still too risky. This comes as no surprise since AI systems remain premature in their abilities to account for all the variabilities in these operating conditions; however, autonomous vehicles do flourish in industrial contexts where operating conditions are routine and predictable.

We also can foresee the expected step-wise adoption of autonomous vehicles in additional industries, where the technology will enter sectors determined by small increments in the complexity of the path followed by the vehicles. Confidence is another take home message from this example. Though criticisms of autonomous vehicles for the consumer market are legitimate, you should not conclude outright that such forms of AI automation is premature for your industrial operations. The same issues are surfacing in other industries where the operational context is variable. This includes IOT, drones, robotic assembly lines, aerospace, defense, medicine and the environment. A good example of an industry that was a late entrant to the AI space is construction.

Construction is catching up

The reputation of the construction industry as a technology laggard is ending–and with good reason. In Canada, the sector is grappling with severe labour shortages and supply chain disruptions due to the pandemic. These are but a few factors that cause construction costs to soar, which Statistics Canada estimated last year to have inflated the price for non-residential projects by 8.3% and residential projects by 20.3%. Since governments are investing heavily in infrastructure projects to stimulate economic development, constraints on supplies of labour and building materials will remain a long-term challenge.

The need for AI automation to cut costs through efficiency gains and reduced labour requirements is obvious. The industry’s focus is now on determining repetitive, well-documented tasks in construction workflows. By partnering with the leading Canadian construction company Pomerleau, we completed a POC project to automate one vital, labour-intensive task: evaluating preconstruction planning documents and engineering specifications. Before striking ground, plans for complex construction projects undergo routine readjustments and changes–sometimes more than 20 times a day. The constant need to adjust orders with suppliers of building materials cuts into the time project manager have to mitigate risks and provide up-to-date information for accurate estimates of project costs.

Construction companies are fortunate in that images of construction plans are plentiful. Associating the amount of building materials, like windows and piping, in plans with cost estimates is also straightforward. Our partnership used these images to develop an AI computer vision solution capable of automated counting of building materials in pdf files of blueprints and engineering specifications. What should grab your attention here is that this cutting-edge AI automation came to light using data that is readily available in this industry. If late-adopters of AI have valuable sources of data at their fingertips, chances are the same is true for your business.

Now is the time for accelerated AI adoption in Canadian industry

Adopting new, unfamiliar technology within critical industrial operations does come with risks. For one, the abstract depictions of AI technology make it difficult for business-leaders to understand and evaluate the benefits and drawbacks of AI innovations, as well as justify their costs. This, too, is changing since AI solution providers like ours now provide visual and intuitive displays of AI automation that are comprehensible to the business community. With our platform, non-AI experts can communicate with AI specialists and challenge a proposed solution, question the assumptions and understand how their data influences AI algorithms. Ultimately, this ensures that end-users can bring real-world operational context to proposed AI solutions and significantly de-risk AI solutions prior to their deployment.

Note that one benefit from the pandemic is that provincial and federal governments continue to funnel much-needed funds towards subsidizing technology adoption by Canadian industry as a means to stimulate economic growth. Be sure to check out the opportunities on offer by SCALE AI Canada, CNRC-IRAP; and if you’re in Quebec, the Ministère de l’Économie et de l’Innovation (MEI) and Investissement Québec. Our case with Pomerleau is exemplary. This successful POC project enabled our partnership to secure subsidies from the leading funding organization INVEST-AI, which helped cover a significant proportion of our R&D efforts. We are confident your business can follow our lead and acquire cutting-edge, trustworthy automation without breaking the bank.