https://www.humanresourcestoday.com/frs/24235077/the-skynet-effect--how-hr-can-best-utilize-ai/email
AI this, AI that. No matter where you go, AI seems to be all anyone in HR wants to talk about. It might be a little irritating, but it’s inescapable for a good reason. Artificial intelligence, specifically ChatGPT, is now an important topic of conversation for all industries.
Like anything new, there are plenty of questions and misconceptions about how AI will change workplace dynamics. But modern AI isn’t Skynet trying to take over the world, and instead of fearing it, your organization can embrace the efficiencies and positive impacts that it offers.
In this webinar, Iveta Brigis, Vice President, People Operations, and Wesley Pasfield, Head of Data Science will address:
• How to better understand AI, specifically ChatGPT
• Use cases for AI in the employer space
• Considerations and questions to ask when evaluating an AI vendor
• Concerns about AI affecting HR employment availability
• How AI can enhance HR jobs when used responsibly
2. Chronic condition prevention and
management
2
Reimagining healthcare to be more scalable,
engaging and proactive.
Conversational member-facing application leveraging
Artificial Intelligence coaching and connected devices to
create personalized experiences
Available anytime, anywhere for all members
Supporting all lines of business
Solutions to drive:
What is Lark?
Care
Activation
Digital companion to clinician-guided
support for sustained weight
management
Weight
Management
Proprietary and Confidential
3. Trusted Partner
Created with leading industry experts, Lark has deep experience working with some of the leading employers
and health plans in the country and has been recognized across the industry.
Baba Shiv, PhD, MBA
Behavioral Economics Expert at
Stanford University
Bob Gabbay, MD, PhD
Chief Scientific and Chief Medical Officer for the
American Diabetes Association
Jo Solet, PhD
Cognitive Behavioral Therapy Expert at Harvard
Medical School
Jonathan Fielding, MD, MPH, MBA
Distinguished Professor of Health Policy at the UCLA
Fielding School of Public Health
Leading Experts
7YEARS
IN R&D
EMPLOYERS
2000+
CDC Full Recognition
Diabetes Prevention
Program Provider
The World’s First A.I.
Healthcare Provider
NPI #1114438157
Lark is a member of
The American Heart Association
Innovator’s Network
Deep Experience
VC FUNDING
$193M
Industry Recognition
MEMBERS
CONTRACTED
30M+
4. Member Activation Weight Management
What Does Lark Offer?
Lark’s solutions span the spectrum of weight loss to cardio-metabolic risk
factors, from prevention to clinical care to the full range of weight
management.
All Solutions Include
Connected Devices
Wellness
Mass Enrollment
& Engagement
Screening & Early
Disease Detection
Care Activation
Healthy Weight
GLP-1
Companion
Prevention
Heart Health
Diabetes Prevention
Diabetes Care
Hypertension Care
Clinician-Guided
Step Therapy
8. What are these buzzwords?
In short - getting computers to automate
tasks in a human-like fashion. Machine
Learning is a form of artificial intelligence
what is A.I.?
what are Large Language
Models (LLMs)?
Machine learning models trained on huge
amounts of text data that are capable of
generating text by predicting the likelihood of a
word given its context in an inbound text-based
request - a subset of Generative AI
what is Generative A.I.?
A subset of artificial intelligence that uses machine
learning techniques to generate new data or content
that is similar to a given set of input data
what is ChatGPT?
ChatGPT is an example of an LLM and really brought
LLMs into the public domain.
Since its release in November 2022 there have been
numerous other similar public releases, notably Anthropic’s
Claude, Google Bard, Facebook’s open-source LLaMA 1.0
and 2.0 among MANY others - and it hasn’t even been a
year yet!
10. Proprietary & Confidential
HR Administration
▶ Chatbots for Customer/Employee Support
▶ Automated note-taking and documentation
▶ Become an “editor” rather than a “writer”
▶ Generate new ideas: “List 10 non-monetary ideas for employee recognition.”
▶ Policy Writing: “Draft a work-from-home policy for an employee handbook.”
▶ Drafting Training Content: “Draft me an outline for training new managers.”
▶ Data Analytics, including Sentiment Analysis “Tell me the top themes from these
employee comments, written at the 8th grade reading level.”
AI in People Operations
Iveta
11. Talent Acquisition
▶ Opportunities
○ Streamlining time and effort: ““Draft a job description for the Chief People
Officer of a health technology start-up.”
○ Reducing overhead by automating parts of the hiring process
○ Designing branded campaigns
○ Drafting job simulations
○ Bonus: Study what predicts success on the job
▶ Things to watch out for
○ Increasing bias
○ Depersonalizing the candidate experience
○ Ignoring the personality behind the resume, does not work for all industries
○ Emerging regulation (i.e. New York City law)
AI in People Operations
Iveta
12. Opportunities
▶ Healthcare access and quality uniformity
▶ App-based
○ Care can be accessed from anywhere
○ The safety of their own home
○ Removes shame from walking into a weight center/mental health facility.
○ Removes biases present in traditional care settings
▶ With intentional safeguards in place, can help achieve greater health equity
among users
Things to watch out for
▶ Can exacerbate biases found in training data set (validation sets)
▶ Can “hallucinate”
AI in Healthcare
Wesley
13. AI in Healthcare and Benefits
Wesley
Healthcare Access and Equity
24/7 Care Coverage
Access whenever, wherever
Aid in Patient Treatment
As an assistant or a replacement depending on
the situation
14. Proprietary & Confidential
Behavior Change Lifestyle Management Programs
▶ Common Wellness Program Offering
▶ AI Benefits from Lark’s Model (Compared to Human-led Programs)
○ Reach & Accessibility Access to programs anytime, anywhere, 24/7
○ Health Equity Same quality of experience for all members
○ Social Determinants of Health Remove biases from the care experience
○ Administrative Overhead Right-size for all workforce sizes
○ Personalized Coaches learn more about your each interaction
AI in Wellness Programs
Iveta
15. Current events
Model training data is up until
a current point so it won’t
have access to events past
that point in its training data
Irrational Confidence
Some responses are
inaccurate and too confident
These are commonly
referred to as “hallucinations”
Variation in Response
Slight changes in prompt can
lead to different answers
Users can override prompt
instructions - hard to control
Unknowns
Model is only as smart as
training data
● Are there certain topics
that contain
disproportionately
incorrect data?
● Are there inherent biases
in the data?
● Are there ways to
manipulate certain
subjects to spit out
incorrect answers?
Current AI + LLM Limitations
17. Considerations and Questions to Ask
Wesley
► Consultative and collaborative partnership
► Understand how they trained their model
► Ask about the safety features they have incorporated into their model
► Understand what cost-savings or ROIs you would be receiving
compared to human-driven solutions
► Work with legal team | Different Parties
19. Many tasks and some jobs have already been, and will continue to be
replaced by machines/algorithms (and governments need to support
those impacted)
But many other jobs will be created by the opportunities that emerge
from this new technology - think of the entirely new categories of jobs
that exist today that did not 20 years ago
● Telemedicine provider (doctors, nurses)
● Social media influencer
● Blockchain analyst
● Big data scientist
● Uber/Lyft driver
● Drone operator
● Online dating profile writers
● Prompt engineers (2023!)
The real question is:
How do you stay up to date
with the emerging
technologies and future-proof
your own career?
Will AI Take Away Jobs?
Creativity and innovation will become a premium
Iveta
22. What to look for in an AI vendor
Are they leveraging their own models?
Are they building or fine-tuning on top of third party models?
Open-Ended: Leveraging probabilistic model output. We
don’t know for sure what will be returned, but content has a
higher potential to be engaging
Pre-Determined: “Hard-coding” a response. We know exactly
what will be returned, but it’s harder to customize to individual
user needs
Look for partners that are conscious of this trade-off, and
when to inject “probabilistic” experiences to drive more
engagement and value for users, and when to fall back to
“determinism” in situations where we need 100% guarantees
Balancing “Open-Ended” and “Pre-Determined”
Where do they fall in the AI stack?
Consult Internal Experts, Legal & Security!
Are they deliberate about their
evaluation framework?
Consultative and collaborative partners
Can they measure and prove their impact?
This is a rapidly developing space - are they willing to
share knowledge, educate, and collaboratively build
solutions?