Consultancy services take several forms:
- Asset management reliability contracts (c 46% revenues). These are fixed term (typically 12 months) asset care service contracts,
which include elements of remote monitoring and monthly site visits to optimise operational
performance and/or support compliance.
- Projects (c 30% revenues), which generally focus on how to optimise and enhance machine performance
and can often arise as a result of an existing contract. Projects can also provide
support to customers during change management programmes.
- MMP (‘My Maintenance Planner’) (c 4% revenues) subscription revenues associated with
the 53 North software-based maintenance scheduler. Periodic data collection is used
to automate elements of a customer’s machine maintenance scheduling and other work
orders.
- Hardware (c 20% revenues), associated with condition-based or project-based work.
53 North originally developed the cloud-based computerised maintenance management
system (My Maintenance Planner, MMP), which is available to customers. MMP helps customer
sites manage work orders and schedule maintenance to best optimise asset management.
The IntelliAM AI platform generates insights that go well beyond the MMP module. It
uses richer data sets from sensors measuring a range of variables, including vibration,
turbidity, speed, temperature, torque and lubrication, as well as information regarding
energy usage and maintenance plans. By collecting, cleaning and storing this data,
it can then apply machine learning algorithms to the information, producing insights
that conventional asset care contracts are not able to generate. It achieves this
by (i) collecting data in real time, (ii) collecting data and other predictive maintenance
inputs in much greater volume and (iii) cleaning and tagging data extract the best
possible insights. The outcomes produced from the machine learning models are thus
much richer than traditional analysis, in terms of analytical granularity and by being
specific to each customer’s operating infrastructure and working practices. Furthermore,
IntelliAM AI can not only identify alternative (more productive) line configurations
but also, via its consultancy operation, help customers with change management.
The group’s future platform services revenues will be increasingly based around this
IntelliAM AI platform and take several forms:
- Setup: to use the platform, the customer’s plant must be digitally mapped to assess which
sensors are available (often sensors are added as needed) and data collected from
the specific equipment configuration that the customer is using. Any historical data
also has to be collected, cleansed and analysed.
- Access: customers will pay a monthly fee to access the models. This access charge is currently
calculated on a per-site, per-month basis, but many AI-based SaaS solutions are likely
to focus on more usage-based charges in the future.
- Usage subscription revenues are currently associated with the number of variables on which
data is being collected and modelled. We anticipate that only a small proportion of
the c 5,000 potential customer variables that can be monitored will be analysed on
adoption, but that the depth of analysis will increase over time.
The customer requirement
Given (a) the significant investment and (b) the significant cost of down-time, it
is vital that manufacturers ensure that plant and equipment operate at the most efficient
levels possible and maintain quality output. Poor overall equipment effectiveness
(OEE) manifests itself in many ways including equipment failure, slow cycles, process
rejects, idling, planned stops (setup and adjustment) and reduced yield. Equipment
failure is the most challenging to anticipate. In a 1978 report entitled ‘Reliability-centred
Maintenance’, Nowlan and Heap concluded that only c 11% of equipment malfunctions
were age-related (ie somewhat predictable based on usage) whereas c 89% were not.
If such a high proportion of equipment malfunctions are not predictable using duration
of operation, then other factors must be responsible (and therefore monitored) in
order to predict machine failure.
This challenge is applicable across the entire manufacturing economy and represents
a significant addressable market for those offering related services in what is known
as condition-based monitoring. This monitors a range of variables in addition to operating
time/age and will include factors such as vibration, temperature and lubricant analysis.
This asset care philosophy begins as early as system design and continues throughout
the lifecycle of the equipment. It manifests itself in the monitoring of production
lines in a number of different ways:
- Programmable logic controllers (PLCs), which can monitor and record live operational data such as machine productivity,
temperature and start/stop processes, and generate alarms if a machine malfunctions.
- Human-machine-interfaces (HMI), which act as a communication link between individuals operating machines and the
system to oversee and control particular equipment.
- Supervisory control and data acquisition (SCADA) systems, which give operators a holistic view of the state of equipment and infrastructure
running in a plant.
At first sight it might seem that manufacturers’ approach to production line optimisation
is already highly scientific, however the reality is that a significant amount of
operator experience is required to interpret and react to the various inputs (ie to
add context to the data). There are a number of inherent inefficiencies in this approach:
- Data sets are heterogeneous. Production lines are built from equipment from different manufacturers with varying
data architectures. It is challenging to build ‘clean’ data sets that are applicable
across multiple sites.
- Available data sets are small. Sensors that collect data are often limited in number and type. Furthermore, the
data that PLCs generate is often constrained by the manufacturer or customer’s ability
to process and store large data sets. This issue also limits the amount of contextual
data (eg tank levels, waste, line speed) that can be added to the machine-level data
to build up a more comprehensive picture.
- Production lines are bespoke. Two identical pumps could operate at different speeds and flow rates, through different
pipe lengths and handle liquids of different viscosity. This makes standardised approaches
to line optimisation almost impossible.
- Production lines are multi-product. Even within the same customer, a single production line might be used to produce
a number of products with different behavioural characteristics. For example, viscosity
will play an important role in determining bottle filling times for skimmed and full-fat
milk.
- Product lines are complex. In any multi-step manufacturing process, the tendency is to divide the process into
clearly defined sections. This tends to focus on efficiency levels for each individual
section rather than taking a more holistic view. In this way many causalities could
be overlooked.
- Manufacturing conditions vary. Seasonal factors such as ambient temperature could affect both plant performance
and product characteristics requiring more contextual analysis.
- High levels of operator reliability. While predictive maintenance contributes to reliability, it does not affect run-time
productivity (e.g. throughput efficiency, reduced waste streams, improved cleaning
cycles, reduced change over time, supply chain improvements etc). For this, manufacturers
rely on experienced operators to operate machines as efficiently as possible. What
emerges therefore is a generally accepted set of operator guidelines that (experience
has shown) results in above-average levels of productivity by historical standards.
Such insights are only available from experienced operators (who are falling in number
as automation rises). Customers are therefore looking to turn asset optimisation from
an art into a science.
IntelliAM AI’s solution
IntelliAM AI provides a multi-step pathway to AI-enabled manufacturing asset care.
It aims to:
- Form the most comprehensive picture of asset performance in real time, allowing operators
to predict, prevent and react to performance issues efficiently,
- Contextualise this performance data based on the specific operating conditions in
order to optimise (1), and
- Identify contextualised performance patterns across complex systems to create new,
meaningful performance insights.
One analogy to the IntelliAM AI platform might be the role of computers in the game
of chess. At a simple level they can act as guiderails, ensuring that players adhere
to the rules of the game. At a higher level they can point out threats from opponents
and advise players on the strength of one possible move against another. At an advanced
level however, they can suggest moves that might not seem intuitive to a human player,
based as they are on the analysis of the widest possible range of future moves that
go well beyond human analysis. In this way, experience of playing the game is augmented
by the mathematics of the game. In the same way, IntelliAM AI’s solutions have the
potential to convert the ‘art’ of optimising asset productivity (which has historically
relied heavily on the experience of operators) into a science. The science will always
provide far more accurate results when the self-learning algorithms are fed with millions
of relevant data points each week, which is now possible through technology convergence.
Therefore, line performance is improved with the application of this technology.
The starting point for IntelliAM is to create a digital map specific to the process
being analysed. This includes assessing the volume and nature of the data being collected,
as well as the infrastructure required to upload data to the cloud (in this case Azure).
An added challenge is that older equipment might generate data in older formats. This
mapping exercise may lead to additional sensors and/or contextual data being suggested
in order to build the most comprehensive picture. We would expect a site on average
to be capable of monitoring up to 5,000 different parameters, although in the early
stages of adoption customers might focus on only a small subset of these for analysis.
Furthermore, IntelliAM AI is capable of collecting c 500 million cleaned and tagged
data points per line per month, using deep domain expertise to identify relevant data
from the billions of other data points that are generated.
The next phase of implementation is to clean and tag the data before IntelliAM builds
a specific profile for each operating process. The large amount of data available
for processing is both an asset and a liability. It is therefore key that the solution
provider identifies, captures and catalogues the precise data required to generate
meaningful insights. Once built, the models can run in real time, providing a wide
range of insights on the performance of individual manufacturing processes and, most
importantly, relationships between them. Site operators can see and act on the output
from the models via a comprehensive collection of dashboards created with Grafana,
an open-source analytics and visualisation tool.
In summary, output from the IntelliAM AI platform can be used at various levels of
sophistication:
- At a basic level the platform can measure condition, or OEE (expressed as a percentage
of maximum productive potential) at levels of granularity hitherto unachievable. This
can be used to facilitate better maintenance by establishing baseline performance
levels, setting alarms and streamlining maintenance workflow with work order creation.
- At an intermediate level the platform can be used to provide contextualised OEE by
analysing machine performance under different operating scenarios, ensuring optimum
productivity across the widest possible range of conditions.
- At an advanced level the platform can identify relationships between only indirectly
related processes, providing insights that can extend throughout the supply chain.
The platform could, for example, provide recommendations on the speed of container
filling and container sealing processes are related based on content viscosity. There
is also the potential to provide analysis with which management can measure the impact
of different input materials on manufacturing efficiency, as well as developing action
plans for emissions and waste.
It is important to note that much of the data used by the platform is already being
generated by the manufacturing assets, but is often unused as it is difficult to prioritise,
clean and tag. IntelliAM AI brings not only the ability to ingest this data efficiently
but, most importantly, provides the ability to extract value from the data that would
otherwise remain unavailable. This is achieved through powerful multi-factor machine
learning models. Running current data through a model allows the algorithm to identify
and ‘remember’ relationships between data sets that would otherwise not be highlighted
by more superficial analysis. This analysis is then converted into a series of operational
recommendations (for example – in simple terms – a recommendation to slow the speed
of machine A, which will lead to an increase in output of machine B and therefore
lead to an overall increase in production line efficiency).
Competition
Competition comes principally from several sources. Equipment manufacturers offer
the ability to schedule maintenance tasks and generate work orders, and ERP solution
providers are looking to add better maintenance workflow to their existing offerings.
Equipment manufacturers are limited in their ability to take a process-wide view or
to harness the powers of machine learning analysis, while ERP solution providers are
looking to add value, mainly via workflow efficiencies rather than via data analysis.
Furthermore, both solution providers do not generally have the expertise to ensure
the data collected produces meaningful insights..
Another source of competition is from machine learning platforms such as C3 AI. The
challenge for such vendors is again gaining access to manufacturing expertise, which
IntelliAM AI has thanks to established customer relationships. Once again, without
this expertise, analysis of the significant volumes of data created is unlikely produce
meaningful results.
Other competitors include Siemens, which acquired UK-based Senseye in June 2022 and
introduced new generative AI functionality into the latest release of its predictive
maintenance offering in February 2024. Its solution integrates with any asset, system
or data source, using existing data or with newly installed sensors. It focuses on
the metals and mining, pulp and paper, automotive and food and beverage industries,
claiming to reduce unplanned downtime by up to 50%. This focus on improving reliability
overlaps somewhat with IntelliAM AI, but does not address the highest value-added
productivity solutions.
Israel-based Augury is another offering, focusing on ‘machine health’, ‘process health’
and ‘production health’. This business model uses subscription-based predictive maintenance
solutions, based on the supply of sensors and analytical tools to minimise downtime.
The platform focuses on sensor solutions to gather data around specific parameters
into the cloud, which then requires subsequent contextualisation in order to optimise
predictive maintenance output. As such it represents just the first (albeit important)
step in leveraging the power of AI solutions. Disclosed customers include Hills Pet
Nutrition, DuPont, Osem-Nestle, Colgate-Palmolive and Heineken.
IntelliAM AI’s strategy
Educate
As is often the case with disruptive technologies, education is playing a key role
in the early stages of the market’s development. Often a key starting point for IntelliAM
AI is differentiating its productivity offering from simple reliability solutions.
Another challenge is setting customer expectations at the correct level. An overly
pessimistic customer mindset will fail to appreciate the significant upside that a
machine learning-based optimisation strategy can ultimately deliver. Conversely, an
overly optimistic mindset could set customer expectations unrealistically high. Early
discussions with customers at a plant level also need to emphasise that AI is not
synonymous with automation. The solution enhances performance primarily via operator
interface recommendations, and not necessarily through automation.
Sell
The 53 North consultancy business forms the core of the group’s customer-facing employees
and brings with it a well-established manufacturing customer base. The ability to
harness the power of AI is a high priority for all businesses, and manufacturing is
no different. The strategic value of AI investment means that the group’s sales strategy
is to develop its top-down, investment-led strategic messaging for C-level management
so that it will resonate alongside its already well-established reputation for providing
bottom-up, problem-led solutions. Alongside this will be the development of the group’s
account management and customer success capabilities as it develops a SaaS revenue
model.
Convert
The group’s consultancy business currently has asset care service contracts covering
over 150 customer sites, including some of the largest global FMCG manufacturers (see
Exhibit 2). Adoption of the IntelliAM AI platform by these customers creates new,
real-time remote data links between IntelliAM AI and the customer, and gives the customer
access to the group’s machine learning modelling. We estimate that the majority of
the group’s condition-based contracts are due for renewal over the next 12 months
(typically December to March), and this represents a significant opportunity for IntelliAM
AI to start transitioning customers onto entry-level AI-based solutions.
Land & expand
Given the heterogeneous nature of customers’ asset configurations, IntelliAM AI continues
to develop models for new applications (eg supply chain), and in the process continues
to expand its reach beyond the FMCG sector. In addition, further development is required
to offer complete end-to-end solutions for customers. Given the strategic nature of
the IntelliAM solution, the group is embarking on a more aggressive ‘land & expand’
strategy over the next few years with the emphasis on establishing a presence across
the broadest possible range of customers before then focusing on building wallet share.
Approximately £3.3m of the £5.1m (gross) raised at the IPO on AQSE will be invested
in hiring additional key personnel in software engineering, data science and automation
engineering. Expansion opportunities also present themselves geographically, given
that the group already counts half of the world’s largest food and beverage manufacturers
among its customers.
Partner
The group has already established strong working relationships with a number of original
equipment manufacturers (OEMs), not least the leading manufacturer of bearings, SKF.
SKF is investigating the opportunities of using IntelliAM AI models to further refine
lubrication regimes as part of an enhanced asset care strategy. IntelliAM AI recently
announced that SKF has signed a letter of intent, which includes the future incorporation
of the IntelliAM AI platform into SKF’s own AI solutions. In July the group announced
a Digital Innovation Fund (DIF) Lighthouse Funding Award of £263,000 for research
into the application of AI in lubrication analysis. This will introduce machine learning
solutions to a wider group of small and medium-sized enterprises (SMEs) as well as
underpin the group’s new product development.
Differentiate
The ability to differentiate the offering in a nascent market is key. Neither asset
care strategies nor many of the necessary sensors are new, and as such many third
parties are quick to label any form of data-based predictive analysis as AI-based.
IntelliAM AI continues to focus on its machine-learning models as well as (most importantly)
the data lake that such models rely on. It is important for customers to understand
the incremental added value that comes from identifying, cleaning and tagging the
relevant information, so that meaningful insights can be gleaned from otherwise meaningless
streams of data – something that is only possible with deep domain expertise. Furthermore,
once line optimisation strategies have been developed, the same expertise is important
to guide any change management strategies.
Underpin
The group’s technology strategy is to build upon well-established third-party infrastructure.
A key component of this is the data intelligence platform Databricks. By combining
the structure of data warehouses with the flexibility of data lakes, Databricks offers
IntelliAM the ability to generate insights by encompassing the widest possible range
of data types. Furthermore, IntelliAM AI can create secure integration between the
Databricks platform and its cloud storage solution of choice, Azure.
The IntelliAM investment case
We believe that IntelliAM AI’s investment proposition is well-balanced, offering investors
elements of both the growth and the quality of cash flow that underpin any going concern
valuation. Although a number of these points have been covered, it is worth reiterating
them within the framework of an investment case.
Growth
- Customer requirement. We believe that IntelliAM AI addresses a very real customer need given the role that
asset care strategies must play in addressing the current stagnation of UK productivity.
The IntelliAM AI platform has already shown sustainable improvements of around five
percentage points in OEE in one food and beverage project (average industry OEE being
c 55%). Management estimates that such an improvement translated across the entire
UK industry could be worth £4bn in additional revenue capacity. Finally, there is
a growing need for manufacturers to communicate along the supply chain in order to
manage issues such as emissions and waste. This can only be done through the collection,
processing and exchange of data.
- Addressable market. While not quantifying the size of the UK opportunity, it is clear that the group’s
current 150 customer sites represent a very small fraction of food and beverage sites
in the UK. The group’s end user markets are also strategically significant. There
are an estimated 12,500 FMCG businesses in the UK alone, representing around £24bn
of exports to more than 200 countries and representing 17% of UK manufacturing gross
value added (sources: FDF, The Food Foundation, gov.uk). Furthermore, the problems
addressed by IntelliAM AI apply equally across the entire manufacturing sector.
- Market share. While the market for AI-enabled asset care solutions remains nascent, IntelliAM AI
already has UK consultancy relationships with five of the top 10 global food and beverage
groups. This is key to ensure that the group is well represented among the companies
expected to have the resources to be early adopters.
- Scalability. Like many SaaS models, the IntelliAM platform is highly scalable. Furthermore, the
platform allows high-volume remote data capture. This greatly expands the group’s
monitoring capability and, as set-up processes become more standardised, the ability
to scale quickly should also expand significantly.
- Current trading. With the majority of current year renewals for condition-based consulting contracts
only occurring in H2, the recently announced H125 results give only a modest indication
of the likely year-end installed base for the group’s AI platform (we forecast 18
sites at year-end, up 16 y-o-y). That said, a significant consulting contract extension
(in beverages) and a contract win (Hovis) contributed to the board being ‘comfortable
with current market forecasts’, underpinned by first half pro-forma consulting revenues
of £1.5m.
