The Power of Learning Through Experiments in Power Apps Development
#154. Neil Benson takes us on a journey of learning through experimentation. Starting with the historical origins of smallpox inoculation, Neil discusses the importance of conducting experiments to drive innovation and problem-solving.
Sharing personal experiences from his biochemistry studies to his current work in building applications with Power Platform and Dynamics 365, Neil emphasizes the value of agile software development and the significance of learning through short bursts of experimentation.
He provides insights into real-world experiments with AI Builder and form processing models, highlighting the impact of these experiments on improving business processes. Neil also shares his experimentation with AI features in his own content creation and business operations, encouraging listeners to embrace the mindset of a scientist and continue experimenting.
So, tune in to explore the power of learning through experiments and discover how it can accelerate your own journey in building amazing apps.
Free Workshop: 3 Secrets to Building Business Apps with Scrum
Join this free interactive workshop to discover the three secrets to successfully using Scrum to build agile apps so you can deliver projects faster, under budget, have more fun, and get promoted. Join today: https://customery.com/3secrets
Mentioned in this episode:
Subscribe to my new Estimating Business Apps podcast mini-series
Estimating Business Apps is a five-part podcast that will help you quickly, accurately and confidently estimate complex Power Platform and Dynamics 365 apps in minutes. Listen now: https://estimating-business-apps.captivate.fm/listen
G'day and welcome to Amazing Apps.
Speaker:I'm your host, Microsoft MVP, Neil Benson.
Speaker:I'm on a mission to help you master Agile
Speaker:practices and build amazing apps on the
Speaker:Microsoft Power Platform and Dynamics 365.
Speaker:Amazing Apps is the result of my
Speaker:curiosity and experiments with new
Speaker:ways of building amazing business
Speaker:apps and high performing teams.
Speaker:It's full of advice from my guests
Speaker:and examples from some of my work over
Speaker:the last few years leading business
Speaker:applications, teams, and practices.
Speaker:If you enjoy this episode, head
Speaker:over to https://amazingapps.Show
Speaker:for additional resources.
Speaker:You'll find more episodes of Amazing Apps
Speaker:as well as my videos, free workshops,
Speaker:ebooks, and my online training courses.
Speaker:In this episode, I'm going to try
Speaker:and persuade you to continue learning
Speaker:through experimentation as we
Speaker:embark on our AI adoption journey.
Speaker:We're going to be talking
Speaker:a lot about experiments.
Speaker:But let's start with smallpox.
Speaker:When's the last time you or
Speaker:someone you know well had smallpox?
Speaker:I bet it's never, at
Speaker:least I hope it's never.
Speaker:According to the World Health
Speaker:Organization, people have been trying to
Speaker:inoculate themselves against smallpox by
Speaker:exposing themselves to the virus since the
Speaker:15th century, maybe as early as 200 BC.
Speaker:In 1721, Lady Mary Wortley Montagu
Speaker:brought smallpox inoculation to Europe
Speaker:by asking that her two daughters
Speaker:be inoculated against smallpox.
Speaker:That was a practice she
Speaker:had observed in Turkey.
Speaker:By all accounts, she was quite
Speaker:the adventuress of her day.
Speaker:Fifty years later, in 1774, Benjamin
Speaker:Jesty makes another breakthrough.
Speaker:Testing his hypothesis, that
Speaker:infection with cowpox, a bovine
Speaker:virus which can spread to humans,
Speaker:could protect a person from smallpox.
Speaker:He and his family were spared from the
Speaker:smallpox infection that swept through
Speaker:the southwest of England in 1774.
Speaker:Jesty was one of several people
Speaker:thought to have practiced inoculation
Speaker:around this time, but the credit for
Speaker:inventing vaccination is generally
Speaker:given to our next character.
Speaker:Twenty years later, in 1796, a
Speaker:British doctor, Edward Jenner,
Speaker:conducted one of the bravest
Speaker:experiments I've ever heard of.
Speaker:He swabbed the ugly cowpox lesion
Speaker:of a milkmaid and used it to infect
Speaker:an 8 year old boy, James Phipps.
Speaker:If any of you know any 8 year old boys,
Speaker:this isn't a practice I would recommend.
Speaker:Phipps was unwell and suffered a local
Speaker:reaction, but he made a full recovery.
Speaker:So what did Jenner do?
Speaker:Two months later, in July 1776, he tested
Speaker:Phipps resistance by infecting him with
Speaker:matter from a human smallpox lesion.
Speaker:Cowpox in humans results in ugly
Speaker:lesions, often on the hands and arms and
Speaker:face, but it's mild and rarely deadly.
Speaker:Smallpox, however, is far more
Speaker:infectious and in 1776 it often
Speaker:resulted in a slow, painful death.
Speaker:It's reported to have been responsible
Speaker:for between 10 percent and 20 percent
Speaker:of all deaths in the 18th century.
Speaker:Whatever happened to 8 year old Phipps?
Speaker:Well, he remained in good health
Speaker:despite the smallpox exposure.
Speaker:He's considered to be the first
Speaker:person vaccinated against smallpox.
Speaker:And did you know, the word vaccination
Speaker:is derived from vacca, Latin for cow.
Speaker:This is a painting of
Speaker:Benjamin Jesty's cow, Blossom.
Speaker:The most famous cow in the
Speaker:world, at least in 1774.
Speaker:Other experiments since then by
Speaker:scientists have led to vaccinations
Speaker:against over 20 human diseases.
Speaker:Receiving vaccines has become routine
Speaker:for many of us, especially since 2020.
Speaker:Many of us wouldn't be here if our
Speaker:antecedents hadn't been vaccinated
Speaker:against smallpox and other deadly viruses.
Speaker:In 1996, I was studying biochemistry
Speaker:at the University of Edinburgh
Speaker:where I spliced the gene from green
Speaker:fluorescent protein, which is found
Speaker:in the jellyfish, Aquorea victoria,
Speaker:through a bacterial vector into
Speaker:yeast, saccharomyces cerevisiae.
Speaker:According to my professor, our
Speaker:experiments were related to gene
Speaker:targeting and cancer research.
Speaker:Through ultraviolet microscopy, we
Speaker:could see exactly where inside the
Speaker:yeast cell DNA was being expressed and
Speaker:proteins were subsequently located.
Speaker:My goal was just to make
Speaker:glow in the dark beer.
Speaker:Can you imagine traffic
Speaker:cop with a UV torch?
Speaker:Honestly, officer, I
Speaker:haven't been drinking.
Speaker:But I found the conversational
Speaker:skills of baker's yeast to be pretty
Speaker:poor compared to C# developers.
Speaker:So I ended up pursuing a career
Speaker:with the IT crowd instead.
Speaker:But my passion for running
Speaker:experiments hasn't abated.
Speaker:Today, I'm the co founder of SuperWire.
Speaker:ai, a Microsoft partner and independent
Speaker:software vendor building engagement
Speaker:applications for superannuation funds on
Speaker:Power Platform, Dynamics 365, and Azure.
Speaker:I'm also the founder of Customery, an
Speaker:online training provider helping Microsoft
Speaker:teams adopt and master Agile practices.
Speaker:In both businesses, we love
Speaker:learning through experimentation.
Speaker:We start with a hypothesis, run a short
Speaker:experiment to test the hypothesis,
Speaker:review the results, and reassess our
Speaker:hypothesis to improve our knowledge.
Speaker:Instead of learning through experiments,
Speaker:lots of development teams attempt to
Speaker:design everything up front, in the
Speaker:belief that if we could just understand
Speaker:enough at the analysis and design
Speaker:phase, that everything will be alright.
Speaker:If you are analysing your users
Speaker:requirements up front, and designing
Speaker:your solution in advance, you're
Speaker:doing it at the point of peak
Speaker:ignorance, also known as Mount Stupid.
Speaker:At the start of your project,
Speaker:your team knows least about
Speaker:the users and their needs.
Speaker:And your users know least about
Speaker:the application you're building.
Speaker:Instead, if you can defer the requirements
Speaker:analysis until the last possible moment
Speaker:before you need to start developing
Speaker:the feature, you'll have learned a lot
Speaker:more about the requirements by then.
Speaker:Don't spend months analyzing
Speaker:requirements before development starts.
Speaker:Instead, work in short bursts.
Speaker:Keep the users involved in planning your
Speaker:experiments and reviewing the results.
Speaker:Learning through experimentation,
Speaker:working in short increments.
Speaker:Emergent analysis and design.
Speaker:Collaborating with users
Speaker:while building the app.
Speaker:We've got a label for working like this.
Speaker:It's called Agile Software Development.
Speaker:Especially the Scrum framework, which
Speaker:is founded on empiricism, which is
Speaker:the theory that we learn from the
Speaker:experience derived from our senses.
Speaker:That is, complex solutions
Speaker:can't be designed up front.
Speaker:We need to learn through experimentation.
Speaker:Let me give you an example
Speaker:of how we experiment while
Speaker:building Microsoft business apps.
Speaker:One of my teams is currently working
Speaker:for a Queensland government department.
Speaker:They register and monitor the
Speaker:training contracts for Queensland's
Speaker:trainees and apprentices.
Speaker:Every year, they process 90,000
Speaker:expense claims submitted by trainees
Speaker:who have attended an approved
Speaker:training class away from home.
Speaker:63,000 of these claims are PDF forms
Speaker:that are emailed to the department,
Speaker:and 17, 000 are submitted online via
Speaker:a webpage developed 12 years ago.
Speaker:A 12-year-old .NET web app is
Speaker:considered pretty modern by
Speaker:this department's standards.
Speaker:How could we improve the trainees
Speaker:expense claim experience and the
Speaker:department's processing efficiency?
Speaker:The first idea we had was a new
Speaker:mobile-optimized Power Pages site that
Speaker:would connect directly to Dataverse
Speaker:where the trainee data is already stored.
Speaker:We would automatically calculate
Speaker:the distance from the trainee's
Speaker:home to the training location.
Speaker:And we already provide a portal for
Speaker:the training provider to confirm
Speaker:the trainee attended the training.
Speaker:And then we would send the
Speaker:payment to SAP for processing.
Speaker:But the department can't force trainees to
Speaker:use a webpage, and many of them are handed
Speaker:PDF forms by the tutor at the end of the
Speaker:training course, and it's easy for them
Speaker:to get the form approved there and then.
Speaker:Instead, we're going to experiment
Speaker:with the Power Platform's AI builder
Speaker:by training a form processing model to
Speaker:read the PDF expense claim documents,
Speaker:turn them into a digital expense claim
Speaker:record in Dataverse so that we can
Speaker:process most of them automatically.
Speaker:We call this type of work a
Speaker:spike in our product backlog.
Speaker:Like a rock clamors spike.
Speaker:Our spikes allow us to safely explore
Speaker:a new rock face and discover if
Speaker:there is a path towards progress.
Speaker:At the same time, our risk of falling and
Speaker:dying is reduced because we time box the
Speaker:spike and contain it into a fixed amount
Speaker:of effort within our two-week sprint.
Speaker:During the sprint review, we'll report
Speaker:the results of our spike back to our
Speaker:stakeholders and invite their feedback
Speaker:about whether or not to pursue that
Speaker:solution or try another experiment.
Speaker:I remember Frieda, our CRM
Speaker:product owner at the University
Speaker:of New South Wales, wasn't happy
Speaker:that all our spikes went well.
Speaker:If every experiment succeeds and proves
Speaker:your hypothesis, said Frieda, then it's
Speaker:because your experiments were too safe.
Speaker:It's only when half of your spikes
Speaker:fail do you know that you're
Speaker:being bold enough and building an
Speaker:amazing new business application.
Speaker:Our government department is also
Speaker:considering implementing a new business
Speaker:rules engine to replace the 20 year
Speaker:old rules engine that supports the
Speaker:legacy PowerBuilder application
Speaker:we're replacing with Power Apps.
Speaker:When a new training contract is
Speaker:submitted to the department, they need
Speaker:to validate the trainee's details,
Speaker:the employer's details, the workplace
Speaker:location, the contract dates, the training
Speaker:organization, the training course.
Speaker:There are hundreds of validations
Speaker:to perform on each contract, and
Speaker:thousands of rules in the rules engine.
Speaker:Instead of a business rules engine with
Speaker:a fixed set of deterministic rules, could
Speaker:we use AI to validate training contracts?
Speaker:Could we build a model of valid training
Speaker:contracts, then train a co pilot to
Speaker:spot invalid training contracts, and
Speaker:ask it to validate all the new training
Speaker:contracts coming into the department?
Speaker:Arguably, this approach is not actually
Speaker:artificial intelligence, it's machine
Speaker:learning, because the system will be
Speaker:identifying patterns in the contracts
Speaker:provided to it And improving its
Speaker:decision making capability based on
Speaker:our feedback about new contracts.
Speaker:Whatever we call it, I think it's
Speaker:an interesting hypothesis to test.
Speaker:What's the smallest, useful experiment
Speaker:we could conduct to help us advance our
Speaker:knowledge about whether AI, really it's
Speaker:ML, could validate training contracts
Speaker:without a hard coded rules engine?
Speaker:Well, we start Sprint 1 on Monday.
Speaker:If you follow me on LinkedIn or
Speaker:subscribe to my podcast, Amazing
Speaker:Apps, I'll let you know the results.
Speaker:I love building in public.
Speaker:Until then, experiment.
Speaker:Find a hypothesis, run a test,
Speaker:learn from the results, share the
Speaker:outcomes with your stakeholders, or
Speaker:better yet, share them in public.
Speaker:But, please don't experiment on 8
Speaker:year old boys or infect anyone with
Speaker:a deadly disease in your attempts
Speaker:to harness artificial intelligence.
Speaker:Thanks for listening
Speaker:or thanks for watching.
Speaker:I hope you enjoyed this Amazing
Speaker:Apps episode and found it useful.
Speaker:If you want to accelerate your
Speaker:career by building amazing Power
Speaker:Platform and Dynamics 365 apps your
Speaker:stakeholders love, then join me
Speaker:in my free interactive workshop.
Speaker:Inside, I share the three secrets
Speaker:to successfully using Scrum to build
Speaker:agile apps so that you can deliver
Speaker:projects faster, under budget,
Speaker:have more fun, and get promoted.
Speaker:Register today at
https://customery.com/3secrets.
https:You'll also find that link in the episode
https:description, in your podcast player,
https:or in the YouTube video description.
https:Until next time, keep experimenting.
Listen On
