Bringing AI Innovation to your enterprise
With all the buzz around AI, where do you start to bring AI to make
your company
to power innovation by the prescriptive power of AI
I have thinking about the increased demand for innovation amidst the
covid quarantines to offer the technologies needed to thrive in our
remote access and social distanced lives. 2020 has been a testimonial
to human resilience and adaptability. The isolation that comes work
social isolation has led to leadership at an individual level though
many might not see it that way when living alone separated from family
and friends and working and doing all that we define as life, albeit
remotely using a 2D screen.Today reminds me of the early days of the Internet ( link to Dec issue
anchor article about RFP
) when suddenly our works showed promise of
expanding in fun and exciting ways and we did not know about much of
the technologies that was being built. Then came a period of
consolidation and several platforms and layers of technology set in.I believe that we are at that stage of promise of opportunities to
innovate in our businesses and create operational efficiencies and new
revenues. This time we know what comes next and can shape the landscape
of technologies to work for us, on our terms, as a business and as
individuals.What does it take to bring AI innovation to your business? With all the buzz around AI, where do you start to bring AI to make
your company to power innovation by the prescriptive power of AI. Yes
buildings need to be connected for remote access and building managers
will be better off knowing how to manage the reduced personnel that
come into buildings proactively so they can manage the operations and
energy efficiency of buildings.(link to Edge AI article from Nov issue)AI and data innovations have been promised to offer the predictive
power of predictive maintenance of assets, predictive behaviour of
people, and in the predictability of the personalization of this
relation of people to places, things and actions. But AI pilots fail
faster than IoT pilots at 85% of AI proof of concepts stuck without
scaling. Similar to typical enterprise pilots the challenge is in being
able to show ROI. With AI howecer, somehow, business people involved
are optimistic about the AI technology even for failed pilots
They believe that AI should be able to solve their problems and instead
they blame themselves for the performance of the algorithm, lack of
clean data to train the models and not having enough budget to hire
more datascientists.The Growing rift between Business users and data scientistsI believe that the biggest hurdle to get started with AI Innovation in
a company or business unit does not lie in the technology or data or
budgets. The real challenge lies in the growing rift between the data
scientist and business leaders. ( I want to highlight this
sentence as call-out for the article)1 Datascientisr speak the language of AI and define problems as
inputs and outputs to an AI model. For example, a building owner might
want to reduce energy cost by 20% They might do this by
optimizing which machines to use by reducing usage of less
sustainable ones. Another approach could be to see which machines are
used less by people and adapt the machine operations to reduce the
overall cost. Each of these might drive the same business outcome of
reduced costs but it translates to different data science problems. You
ask the AI to find the machine that consumes less energy in the former
and you ask the AI for the machine that is used less by the user in the
former. Both arrive at the same business ourcome but the underlying AI
is a totally different model.2. Data is the language of training the AI. Business users own customer
data, production data or operational data. Data scientists train the
initial model using public datasets based on their assumption of
business needs. They will need business users to bring their business
acumen to translate the context of the company data to improve the AI
models.3. AI algorithms are a black box today. How an AI model makes a
prediction is not transparent to the business user who might be
concerned about trusting a model that challenges many years of their
business experience in their industry. So when an AI recommends to shut
down a turbine how can a business owner with decades of experience
listen to it? The realiry translates to trust between people and the AI
is not questioned but it comes down to the trust on the Datascientist,
trust on themselves to understand the complexity of AI predictions and
the trust that they had contributed to capturing years of business
acumen into an AI model.These reasons cause a disconnect between the business users ask and
data scientist needs to talk the language of data. Then in lies the
failure of 85% of AI pilots not scaling past a proof of concept while
the exuberance on AI solving business problems remains intact.How do you bring AI Innovation to the enterprise ?1. Start with the customer.Focus on solving customer problems and arrive at if and how AI can
solve the problem. Business users need to stay focused on business
problems but need to learn to translate their ask as AI problem
statements.2. Quantify business outcomes.If you want a 50% reduction in churn, you could achieve it even if the
AI algorithm is not fully optimised and performs at a 20% statistical
confidence. If you have a clear quantifiable expectations of
business results, then the business user can improve the AI model in
partnership with the data scientist to exceed expectations.3. Focus on the dataData prep takes 70% time in AI problems. So look at what data is
needed to solve your customer problem and where are the gaps in the
data and clean your data to get good veracity and all desired features
before you start the AI modeling cycle.The way forward to get started with AI innovation is to get the
business user on the same page as the data scientist to become AI
Translators. This is done by a new set of evolving tools and platforms
called NoCodeAI.What is NoCodeAI?NoCodeAI started as a way for the datascientist to automate the AI
model selection process to find the right AI algorithm for a specific
problem, known as AutoML. AutoML stands for automating Machine
Learning. Recently it is evolving to a new field with a platform
for the business user to use drag and drop interfaces to build AI
models. This helps the business user focus on their data and convert
their problem to a data science problem and understand the complexity
of building the model. The business user does not build the model. The
data scientists still builds the models .This frees up the
Datascientist to focus on solving more complicated data science
problems and offer simpler models that the business user can apply on
their data to optimize their data to find solutions for desired
business outcome.We are in early stages of NoCodeAI but it is time for the business user
to understand their role in powering AI Predictions to achieve the
promise of innovation that is inviting us as we manage the covid
reality and get ready to request proposals for the next round of
innovations.Sudha Jamthe is a Technology Futurist with a learning community with
online courses and a live learning lab for product and business users
to bring innovation to their business as they pivot their careers with
NoCodeAI for business managers. www.businessschoolofAI.com
footer
[Home Page]
[The
Automator]
[About]
[
Subscribe
]
[Contact
Us]