salesforce ai presentation

• App Building
• Be Release Ready – Winter ’25
• Integration
• Salesforce Well-Architected ↗
• See all products ↗
• Career Resources
• Salesforce Admin Skills Kit
• Salesforce Admin Enablement Kit
• Career Paths ↗
• Trailblazer Career Marketplace ↗
• Trailblazer Community Online ↗
• Trailblazer Community Group Meetings ↗

Home » Article » How to Prepare Your Org for an AI-Driven Future

• How to Prepare Your Org for an AI-Driven Future

This blog post was written by myself and Tom Bassett and is an abbreviated version of the presentation we delivered at Dreamforce ’23, where we explored how you can prepare your Salesforce org for the artificial intelligence (AI) revolution.

We invite you to read this post, which outlines the key steps you can take, but we also encourage you to watch our full presentation in the video linked below.

The world is changing

With the rapid pace that AI is evolving, how do you make sure you don’t get left behind?

AI runs on data, so the specifics of what is coming in the future of AI doesn’t matter as much as being prepared to feed it properly no matter what comes. The good news is that a lot of what you would do to prepare for an AI future also looks a lot like what you should already be doing now to take care of your Salesforce org.

Admins have historically had a difficult time allocating resources to org and data cleanliness projects because they are often up against the requests of other business stakeholders that are in charge of revenue-generating activities. With the potential revenue and productivity increases that are likely to come with an AI-driven future, admins can now build an even stronger case for prioritizing the activities that will maintain the health of their orgs.

Having a strategy now will prepare your org for an AI future AND still reap massive benefits TODAY.

What’s generative AI?

Before we dive into the Crawl , Walk , Run , and Soar phases of how you can prepare your org, let’s set the scene by explaining what generative AI is and why it’s set to shake things up. This new type of AI can be used to generate content, hence its generative nature. This includes text, translations, images, and answering full-on questions.

Just because you’ve heard of ChatGPT doesn’t mean you understand AI. AI isn’t new, but the pace it’s evolving and reshaping our world is. Salesforce started with AI in 2014, and in 2016 Einstein was born. Over time we’ve seen Einstein Bots, Predictions, and Insights start to change the way we work. And, recently, the fuse has been lit on the rocket with new innovations including Einstein GPT and Slack GPT, which are powered by generative AI.

With AI being on the rocketship, it doesn’t mean you need to launch that massive internal AI initiative immediately. Let’s start off by teaching you how to CRAWL with AI to build up your confidence, become an advocate, and start to think about how it can help you and your business!

A great way to get your head around this technology is to check out the resources on Trailhead. You can start with the introductory modules such as Generative AI Basics and then advance into more complex areas such as Responsible Creation of Artificial Intelligence .

Get AI Associate certified

Once you’ve built a good foundational skill set, why not demonstrate it to the world? In addition to using your Trailblazer Profile to showcase your newly-earned badges, you can become a Certified AI Associate too. The exam guide outlines the costs, passing score, and topics covered. Being an Associate exam, it includes free retakes and doesn’t have any prerequisites—so what are you waiting for?

Follow the trail

Let’s summarize this first phase with a three-step guide on how to master the basics of AI.

Step 1: Get started by following the trail on Trailhead to learn the new terminology so you can confidently know what you’re speaking about.

Step 2: Connect the dots to your business cases, role, learning path, and industry by speaking to your stakeholders and applying what you’ve learned.

Step 3: “Be More Advocate” for AI within your business, and continue to think about how it can save time and money, increase productivity, and support the generation of new ideas.

Once you start feeling comfortable—because you’ve skilled up and your brain is full of cool ideas for how you can apply AI to your business—it’s time to WALK.

Discover and document

Every great admin knows that the beginning of any project calls for them to put on their Business Analyst hat (or T-shirt ). After researching AI and identifying important business processes that could potentially benefit from it the most, you’re going to want to prioritize that list based on potential future business value added, grab the process at the top of your new AI backlog, and get to work documenting it.

Some documentation to create or clean up includes:

• Business processes — Create a business process map.
• Data model — What data in your org connects to this business process?
• Automations — Which automations in your org connect to this business process?
• User stories — Do you have user stories that document why things were built the way that they were?
• Training guides — What documentation was created to teach users this process in Salesforce? Is it up to date?
• How AI is already being used — How are users already leveraging AI tools today to support this business process? And if you don’t think anyone at your company is using AI, do some discovery. You might be surprised.

It’s worth noting that not all of this documentation integrates to AI offerings today, but this is why we’re planning for the future. The idea is to organize to get ahead.

We have the first business process we want to focus on, and we spent time in the Walk stage documenting it and how it connects to Salesforce. When we RUN, we clean house by getting the data ready for AI.

Descriptions

Part of the work in this stage includes updating descriptions in your Salesforce org (and reading the article by Ian Gotts on how to make them AI ready ).

Again, don’t try to add descriptions to everything all at once—that would be overwhelming. Focus on the most important parts of the org that touch the highest-priority business process that would benefit from AI.

Some great places in your org to clean up or add descriptions include:

• Validation rules
• Permission sets
• Lightning pages
• Record types

Descriptions should indicate more than just what the thing is. You also want to write how it works. Why was it created, and why was it created like THAT? How was it derived?

Keep your descriptions concise. Don’t use a bunch of corporate acronyms. Use simple language.

If you have data getting used by AI within Salesforce, it’s also important to add that to those descriptions so if you deprecate that field, you’ll be alerted to update your prompts. Deprecating that field won’t break your org, but your AI would be wrong if your prompts are not also updated.

Make descriptions easier for the AI to zero in on the information it needs to support your business and for yourself to zero in on the information you need to support the AI.

Data cleanup also extends to comments. By adding or cleaning up comments, you’re not only supporting your users, Salesforce team, and consultants but also helping your future AI.

Some areas to review here include:

• Guidance for Success
• In-App Guidance

These will support your users, staff, AND your AI. It’s part of the team now.

Establish data hygiene

Let’s talk about data hygiene. The fact of the matter is if you put garbage data into AI, you’ll get garbage out. Let’s try to establish good hygiene where our data is concerned so it’s up to date and someone is responsible for it.

Within Salesforce, make sure you squash duplicates to reduce the amount of noise you feed into your AI model. There’s no need to keep all of those test accounts anymore!

Review and uninstall packages that are no longer required during a period of spring cleaning, like that package you installed but never got around to using.

Look over your fields and objects and deprecate unused elements to avoid creating a biased AI. Also, take the time to ask yourself, “Are you capturing the information you need to power your AI model?”

Start with key fields and business processes first to break this process down into manageable chunks.

We’ve learned to Crawl, Walk, and Run, now let’s get ready to SOAR through the final phase!

Be well architected

Don’t let the word architect frighten you away from the Well-Architected framework . From admin to architect, we each have our role to play. If you take a shaky architectural foundation and then try to lay AI on top of it, your AI won’t deliver as much business value.

Ensure your system is Trusted in that it’s secure , you are complying with any legal regulations, and it’s reliable for your users, customers, and partners. For example, a secure system isn’t one where all users can export data and feed it into public AI models.

Think about Easy as in it has clear intentions to help users perform Jobs to be Done, while helping them out with automation and keeping them engaged .

You and your system need to be Adaptable to the change that AI will bring by building a system that is not only resilient but also composable , meaning it’s made of interchangeable blocks. You can swap out these blocks as AI changes or your business evolves as a result.

Implement governance strategy

The second part of soaring is implementing governance. We recommend a document, centralize, and evangelize strategy. Document the standards, put them in an easily accessible, central location so everyone knows where to find them, and then let everyone know that those are the standards going forward. By doing it this way, if people fall back into bad habits you can point out, “HEY, HEY… how this should be done is documented right here. There’s no excuse for not doing it right.”

You need governance for your data. You can’t just clean it up once—you must proactively work to keep it clean. Make sure you have a solid backup in place too.

You also need governance around development. Taking care of a Salesforce org is a team sport. Even you solo admins know that someday you’re going to have to call that consultant in for a one-off project, or the team will need to be expanded. Start putting your governance in place so that the org can be guided by the same principles no matter who is doing the configuration on it. Consider things like “clicks, not code,” documenting guidance on naming conventions, and maturing your DevOps process.

Lastly, you’ll want AI governance. You, as the keeper of the company’s most vital data, need to be part of that AI conversation within your organization. Again, Be More Advocate. Consider security and how AI is used outside of Salesforce, and document best practices on prompting and how those prompts will be monitored if they’re used in the org.

Clean house, set the rules, and then AI will come in and make it better.

Your action plan

To round things off, let’s leave you with an action plan that you can use beginning next week to prepare your own Salesforce orgs for an AI-driven future.

• Starting Monday , you’re going to skill up. Start building your AI vocabulary. Think about how AI might benefit your organization in the future. Talk to the people in your organization to see how they’re using AI today.
• Next week (while still skilling up, of course, because #AlwaysBeLearning), you’re going to identify the high-priority business processes that can benefit from AI and put them in priority order. Build and prioritize that AI backlog.
• Next month , take the business process at the top of your backlog and document it. Start to see where that process connects with your Salesforce org.
• For the next 6 months , focus on that high-priority business process and ensure any technical debt is cleaned up and that data hygiene doesn’t leave you guessing.
• For the 6 months after that , establish data, development, and AI governance so that best practices are embedded in the organization and everything you do.
• This shifts into becoming an iterative process with Ongoing Innovation that you can now use to ‘rinse and repeat’ for other lower-priority business processes.

This action plan is packed, but it will help break your prep down into manageable tasks so that you feel confident incorporating AI into your business, no matter what lies ahead.

• External Site: Elements.cloud: MDD (Metadata Description Definition): AI is reading your descriptions
• External Site: YouTube: Be More… Episodes
• Trailhead: Get Started with Artificial Intelligence
• Salesforce Architects Site: Overview: Salesforce Well-Architected Framework

Vanessa Grant

Vanessa Grant is the Salesforce Product Owner at FinTech company Mosaic and the co-host of the Salesforce Career Show podcast. She has built her career around solving business problems with technical solutions and process improvements. She is committed to improving outcomes for all Salesforce projects by speaking regularly on the importance of quality business analysis, devops, and design.

• The Importance of Business Analysis as a Salesforce Admin

Related Posts

Harness Einstein Copilot Analytics and AI for Organizational Success

By Itay Oren | May 8, 2024

As a Salesforce Admin, you know that the rollout of any new solution comes with a few stock questions in tow. No matter how good it feels to accomplish quick tasks with artificial intelligence (AI), eventually someone will ask, “How much is this really helping our company?” As a component of Salesforce’s Einstein Platform, Einstein […]

How to Create a Continuous Optimization Loop with Salesforce Einstein and Data Cloud

By Manjeet Singh | April 29, 2024

As a Salesforce Admin, you know how important it is to get timely feedback on solutions your users rely on. When it comes to solutions like Einstein Copilot—where the user interface (UI) is pervasive, operates in real time, and might give slightly different answers to different users—how can you stay ahead on the robustness of […]

The Ultimate Guide to Prompt Builder | Spring ’24

By Raveesh Raina | February 6, 2024

Artificial intelligence (AI) is not a new concept to Salesforce or to Salesforce Admins. Over the years, Salesforce has empowered admins with a user-friendly interface for the setup and configuration of predictive AI features such as Opportunity Scoring, Lead Scoring, Einstein Bots, and more. The introduction of generative AI in Salesforce brings even more possibilities […]

Diagram Scope

This roadmap shows the components that will be rolled out for AI across the Salesforce Customer 360.

• AppAssessor

Articles by role:

• Consultants
• Business Analysts

By Lucy Mazalon

By Ben McCarthy

• Courses Login

What’s trending

The Rise of Offshoring in the Salesforce Ecosystem

Explore REST API With Salesforce Inspector Reloaded

Salesforce Winter ‘25 Treasure Hunt: Preview Orgs Are Live!

Top 5 Coding Languages Used by Salesforce Developers

Top 19 Salesforce YouTube Channels

UPCOMING EVENTS

How entry-level salesforce professionals can use data cloud to land a job, montréal dreamin’.

Montreal, Canada

How Are Salesforce Promoting the Use of AI [Interview]

Mastering sales order processing in salesforce, salesforce commerce masterclass: a beginners workshop to retail solutions, breaking news: salesforce announces ai cloud for einstein gpt.

The main announcement: Salesforce AI Cloud is claimed to be the fastest and most trusted way for Salesforce customers to deliver generative experiences across all their applications and workflows (which we’ll explain later). 

Salesforce AI Cloud

• Sales: AI Cloud will enable sales reps to quickly auto-generate personalized emails tailored to their customers’ needs. 
• Service: Auto-generate personalized agent chat replies and case summaries.
• Marketers: Auto-generate personalized content to engage customers and prospects across email, mobile, web, and advertising. 
• Commerce: Merchandisers can auto-generate insights and recommendations to deliver customized commerce experiences at every step of the buyer’s journey.
• Developers: Auto-generate code, predict potential bugs in code, and suggest fixes.

Einstein GPT Trust Layer 

Generative AI Models + Salesforce: Your Choice

• Salesforce’s private AI models: In other words, those developed by Salesforce, which are leveraged by customers using features from the Einstein technology layer.
• Combine OpenAI’s ChatGPT: Thanks to the Salesforce and OpenAI partnership, providing out-of-the-box generative AI capabilities to companies through direct access to OpenAI’s enterprise-grade ChatGPT technology.
• Combine Vertex AI models: Thanks to the new partnership between Salesforce and Google Cloud , you can build AI models in Vertex AI, then bring them into Salesforce to be used alongside Salesforce data. Vertex AI is a popular choice as it enables analysts to manage machine learning (ML) projects throughout the entire development life cycle, including training, testing, monitoring, deploying, and governing models at scale.
• Bring your own model: As Einstein GPT is open and extensible, you can opt to use it alongside your own external model(s). 
• 73% of employees believe generative AI introduces new security risks.
• Nearly 60% of those who plan to use the technology don’t know how to keep data secure.

Lucy Mazalon

Lucy is the Operations Director at Salesforce Ben. She is a 10x certified Marketing Champion and founder of The DRIP .

More like this:

Overcoming limitations of the salesforce lead object.

By Timo Kovala

Salesforce Data Migration Guide for Admins

By Tom Bassett

6 Ways to Make Salesforce List Views Run Faster

By Christine Marshall

Leave a Reply Cancel reply

Save my name and email in this browser for the next time I comment.

AI NOW Tour

Get hands-on with Einstein AI and Data Cloud

About AI NOW

Generative AI is revolutionizing app development, creating the biggest opportunity for developers and admins in decades. Join us for a free intensive full-day workshop to get hands-on with Data Cloud, Einstein Copilot, and Prompt Builder. Learn directly from Salesforce engineers and Developer Advocates and get your AI and Data questions answered. Leave with the skills you need to build the next generation of generative AI applications on the Einstein 1 Platform.

Who should attend?

• Administrators
• IT Managers
• Consultants
• Business Analysts

About 5 mins

Learning Objectives

Ai in the spotlight, gpt’s predictive power, a future with generative ai.

• Challenge +100 points

Get Started with Einstein Generative AI

After completing this unit, you’ll be able to:

• Describe the core capabilities of generative AI.
• Describe how Einstein incorporates the power of generative AI.

Over the past decade we've witnessed artificial intelligence (AI) steadily improve at doing things we once thought only humans could do. This has led us to imagine a futuristic world with AI, but one that's still years away. Or so we thought. In early 2023 ChatGPT burst onto the scene, showing the world that the future is already here. Suddenly, AI is responding to complex questions with equally complex answers that sound as if they were written by a human. And because ChatGPT was released for anyone to use, people experienced first-hand how well it could hold a natural-sounding conversation. It has surprised and delighted, and sparked our imaginations yet again. It’s also raised questions about what it means for the future of work.

In this Quick Look badge, you learn about GPT and how it fits into the larger concept of generative AI . Then, you'll learn how Einstein is the world’s first generative AI for CRM.

Okay, let's get this out of the way: GPT stands for Generative Pre-trained Transformer, which means a lot to AI researchers but not so much to you and me. Let's just stick with GPT and focus on what it does.

At a high level, GPT is an AI tool that has been trained to write like we humans do. All it needs is a starting point in the form of a question, request, or even just the beginning of a sentence. And then it continues the thread to create paragraphs of text that are similar to something a real person would write. ChatGPT is a version of GPT that has been fine-tuned to better understand the intention of our instructions, and respond accordingly. You can even tell it to write in a certain style, whether it's poetry, a programming language, or as a pirate. For example, let's see what happens when you ask it to summarize the moon landing in pirate-speak:

"Ahoy there! Avast ye, matey! 'Tis a tale o' adventure on the high seas o' space! Back in the year o' our Lord nineteen hundred and sixty-nine, a band o' brave sailors from the land o' America set sail on a quest to reach the moon. They be Neil Armstrong, Buzz Aldrin, and Michael Collins, and they be on a mission from NASA to land on the moon and plant the stars and stripes flag."

It continues from there, completing the tale in the same silly vernacular.

So how does it generate such convincing text? OpenAI, the creators of GPT, used a ridiculous amount of computing power to train it. They had it read millions of web pages to learn what words are used near each other, and how they're used together. That made GPT really good at predicting the next word of a sentence by using context. So, imagine you're asked to fill in the blank: "In conversations, my partner and I finish each other's _______." You're likely to choose the word "sentences," and so is GPT. But if you change the context a little: "In restaurants, my partner and I finish each other's _______." There's a better chance you would fill the blank with "sandwiches." Same with GPT.

GPT writes paragraphs that follow a thread by making an educated guess at each new word in a sentence. It's way better than your phone's predictive text, which if used in the same way only gives you bizarre sentences that make for funny internet memes.

GPT's ability to write coherent text falls into a broader category of AI known as generative AI . This includes AI tools that can create things like images or audio. As computing power increases, and research continues, generative AI will get better at creating work that’s convincingly human-made. It probably won’t be long before generative AI is part of your everyday life.

The arrival of mainstream generative AI has raised a lot of questions about the future of work. People wonder if they'll be replaced with AI. They're concerned about the ethical use of a tool that essentially impersonates humans. They’re skeptical about the quality of what generative AI produces. Obviously, AI must be handled with care.

Salesforce knows it's critical to establish principles for the ethical use of AI technology. And for the last 6 years, we've had an AI ethics group developing these principles. So when we launch new AI technology, we can help our customers navigate these difficult waters to do the business of business smarter and fairer. Check out the resources section to learn more.

Our latest AI technology brings the incredible power of generative AI to Einstein and the Salesforce Platform.

Einstein is built right into Salesforce, so it can use your own data to custom-tailor everything it generates to your unique business. And since Einstein is available across the entire Einstein 1 platform, it's there to improve every customer experience.

We also recognize that the world of generative AI includes more than just ChatGPT. So we've designed Einstein to allow connections to other language models, such as Anthropic's Claude. This lets you "bring your own model" thanks to APIs and an open ecosystem of generative AI partners.

Einstein is here to empower your business with amazing generative AI capabilities, using your data and your models for your experiences.

• Salesforce Research: Trusted AI
• Salesforce: How Salesforce Research Uses Generative AI as a Force for Good
• Salesforce Help: Einstein Generative AI Glossary of Terms
• Salesforce Help: Einstein Generative AI
• Blog: How to Accelerate Your Success in the Age of AI
• Get personalized recommendations for your career goals
• Practice your skills with hands-on challenges and quizzes
• Track and share your progress with employers
• Connect to mentorship and career opportunities

Site Search

• 1-888-747-9736
• Request a Call

Other Logins Marketing Cloud, Trailhead...

Marketing Cloud

Advertising Studio

Social Studio

Press Release Details

Mobile site search, investor relations, salesforce announces timing of its second quarter fiscal 2025 results conference call.

Results to be released on Aug. 28, 2024, after market close

SAN FRANCISCO--(BUSINESS WIRE)-- Salesforce (NYSE: CRM), the #1 AI CRM, today announced that its second quarter fiscal year 2025 results will be released on Wednesday, Aug. 28, 2024, after the close of the market. The company will host a conference call at 2:00 p.m. (PT) / 5:00 p.m. (ET) to discuss its financial results with the investment community. A live webcast and replay of the event will be available on the Salesforce Investor Relations website at www.salesforce.com/investor .

About Salesforce

Salesforce is the #1 AI CRM, empowering companies to connect with their customers in a whole new way through the power of CRM + AI + Data + Trust on one unified platform: Einstein 1. For more information visit: www.salesforce.com

Mike Spencer Salesforce Investor Relations [email protected]

Carolyn Guss Salesforce Public Relations [email protected]

Questions? We’ll put you on the right path.

Send any questions and requests our way.

Or, connect with Investor Relations at 1-415-536-6250

New to Salesforce?

• What is CRM?
• What is Salesforce?
• Help Desk Software
• Marketing Automation Software
• Explore All Products
• What is Cloud Computing?
• Customer Success
• Product Pricing
• Subscribe to Salesforce
• Salesforce.org
• Sustainability
• GDPR Readiness

Popular Links

• New Release Features
• Salesforce Mobile
• AppExchange
• CRM Software
• Salesforce LIVE
• Salesforce for Startups
• América Latina (Español)
• Brasil (Português)
• Canada (English)
• Canada (Français)
• United States (English)

Europe, Middle East, and Africa

• España (Español)
• Deutschland (Deutsch)
• France (Français)
• Italia (Italiano)
• Nederland (Nederlands)
• Sverige (Svenska)
• United Kingdom (English)
• All other countries (English)

Asia Pacific

• Australia and New Zealand (English)
• India (English)
• ประเทศไทย (ไทย)
• Terms of Service
• Responsible Disclosure

• Content collaboration

Getty Images/iStockphoto

Salesforce, Box integrate document AI into Slack

As salesforce prepares to roll out its own generative ai tools for slack this fall, box adds ai for joint users, who can use genai on their box content in slack channels..

• Don Fluckinger, Senior News Writer

Joint users of Box and Slack can invoke Box generative AI tools on documents shared in Slack channels as of today.

The news comes ahead of Dreamforce, where Salesforce is expected to announce Einstein Copilot for Slack, which Salesforce previewed in June at its World Tour Boston event. Copilot for Slack will have some overlapping features for document summarization, but Box AI will unlock content within enterprise Box repositories that have security and access permissions to protect sensitive content -- and will keep it out of the hands of workers who don't.

Last month, Box AI moved from a consumption model that gave users a set number of queries to unlimited queries for Enterprise Plus subscribers. Box AI can access a number of competing large language models , including Google Gemini and OpenAI's GPT.

Box AI and Einstein Copilot for Slack are also not to be confused with Slack AI, generative AI features released earlier this year that summarize threads and enable Slack users to ask questions about information contained within them.

Features beyond document summaries also going live today in the Box-Slack AI integration include the following:

• Secure file previews of the most recent versions of entire Box documents, including PDFs, spreadsheets and slide decks from within Slack.
• Content cards that show in Slack changes made to a file as the Box file itself is changed.
• Updated Box for Slack workflows, which enable Slack users to create no-code workflows that incorporate Box content and automate business processes that tap Box documents for data.
• Admin tools for joint user authentication.

Salesforce has been touting its Einstein Copilot and Data Cloud products as technologies that can pull business value out of unstructured data, which is a difficult promise on which to deliver technically, said Alan Pelz-Sharpe, founder of Deep Analysis. Adding Box AI within Slack is one step closer to accomplishing that.

Box could become one of Salesforce's most important integrations. When you consider that Salesforce has around 11,000 partners, that's saying something. Alan Pelz-Sharpe Founder, Deep Analysis

"Box provides Salesforce with the ability to secure, collaborate with and manage files to a degree that Salesforce alone cannot provide to its customers," Pelz-Sharpe said. "Box could become one of Salesforce's most important integrations. When you consider that Salesforce has around 11,000 partners, that's saying something."

In Boston last month, Slack CEO Denise Dresser said Einstein Copilot for Slack was planned for release in September. Generative AI will bring customer data and conversations in Salesforce into Slack channels -- and also bring Slack conversations into the Salesforce CRM.

"You [will be able] to talk to your CRM like you would a co-worker or colleague with Copilot in Slack," she said, "bringing all of that customer contact, all of that customer data right where you're interacting and moving work forward."

Don Fluckinger is a senior news writer for TechTarget Editorial. He covers customer experience, digital experience management and end-user computing. Got a tip? Email him .

Related Resources

• 10 Content Collaboration Platforms for Enterprises in 2024 –TechTarget
• A Computer Weekly buyer's guide to the future of collaboration –TechTarget ComputerWeekly.com
• Computer Weekly – 9 October 2018: NHS app marks a new era for UK health service –TechTarget ComputerWeekly.com

Dig Deeper on Content collaboration

Tableau launches generative AI assistant, updates platform

Tableau CEO on AI moves to ease analytics adoption

Salesforce to roll out marketing, e-commerce Einstein Copilots

Salesforce adds generative AI tools to Education Cloud

AR and VR data visualizations offer a new perspective to capture patterns and trends in complex data sets that traditional data ...

Data discovery can use sampling, profiling, visualizations or data mining to extract insights from data. Choose from 10 of the ...

Data professionals can use LLMs for data and predictive analytics work. Still, the analysis of large amounts of textual and ...

Vector databases excel in different areas of vector searches, including sophisticated text and visual options. Choose the ...

Generative AI creates new opportunities for how organizations use data. Strong data governance is necessary to build trust in the...

Snowpark Container Services aims to provide the vendor's users with a secure environment for deploying and managing models and ...

A 3PL with experience working with supply chain partners and expertise in returns can help simplify a company's operations. Learn...

Neglecting enterprise asset management can lead to higher equipment costs and delayed operations. Learn more about EAM software ...

Inventory management software can help companies manage their supply chain in a turbulent world, but using the right product is ...

With its Cerner acquisition, Oracle sets its sights on creating a national, anonymized patient database -- a road filled with ...

Oracle plans to acquire Cerner in a deal valued at about $30B. The second-largest EHR vendor in the U.S. could inject new life ...

The Supreme Court ruled 6-2 that Java APIs used in Android phones are not subject to American copyright law, ending a ...

As SAP pushes its clean core methodology for S/4HANA Cloud environments, the partners who customized legacy SAP systems will need...

Two executive board members will depart SAP in a move that the company says is both to streamline the structure of the board and ...

Sophia Mendelsohn talks about SAP's ambitions to both set an example of sustainability and be an enabler of it with products such...

More From Forbes

The secret to successful enterprise ai ‘human-in-the-loop’ design.

• Share to Facebook
• Share to Twitter
• Share to Linkedin

Carter Busse, CIO at Workato .

After a recent presentation, a technology leader in the U.K. asked me: Since we use AI in so many of our processes at Workato, how do we control it? It’s a fair question. The top concern many have when using AI is inaccuracy, according to McKinsey research.

My answer: human-in-the-loop design.

For many in the tech industry, it’s clear that generative AI is transformative. However, the current large language models (LLMs) are an intermediate step to that transformation. To make it work in the meantime, human-in-the-loop design is emerging as a popular trend to both deploy AI in the real world today and deliver some of the most effective AI-powered business processes.

The Importance Of Human-In-The-Loop Design

I’ve noticed several high-profile examples of AI processes that would likely have benefited from a human in the loop. For example, Air Canada was recently found to be financially responsible after its chatbot mistakenly provided a discount that went against official company policy. Mcdonald's canned its drive-thru AI for serving up bacon on ice cream and placing hundred-dollar McNugget orders.

I think there would likely be many more examples if businesses weren’t too nervous to actually release and scale these systems in the real world. Using LLMs alone, many still struggle to prevent inaccurate or misleading outputs. As a result, projects remain confined to the experimentation stage. That’s preventing organizations from getting the full value from their AI investments.

Apple iPhone 16 Pro To Gain Most From Powerful Upgrade, New Leak Says

Wwe raw results: winners and grades on august 5, 2024, today’s nyt mini crossword clues and answers for tuesday, august 6th.

Already, AI can immediately improve some common tasks without the need for human-in-the-loop design. The technology excels at summarizing huge amounts of information, which is helping professionals in all roles get more immediate access to intelligence. But eventually, LLMs will become so intelligent that they will begin to function as “digital brains” across the enterprise . They’ll replicate human capabilities and drive new levels of automation.

Many businesses are years away from that capability, but they can start to work toward that goal by introducing AI at a greater velocity—and keeping humans involved at critical junctures in the process.

Taking The Time To Review

Recent research by MIT and Accenture shows that, when nudged to review LLM-generated outputs, humans are more likely to discover and fix errors.

In the study, participants were broken down into three groups. Each received the same ChatGPT-generated passages but were provided different levels of insights into what was correct, incorrect or missing from the output. The researchers also included unlabeled inaccuracies throughout the results provided to everyone to further test the ability of users to evaluate AI-generated content.

Those who were given no details about the accuracy or sourcing of the information—the typical experience for most commercial LLM users today—were more likely to overlook mistakes or inaccuracies compared to those provided with error labeling. What this suggests is that, when given the chance to evaluate results from AI systems, users can greatly improve the quality of the outputs. The more information provided about the origins and accuracy of the results, the better the users are at detecting problems.

This is why human-in-the-loop design is so powerful. It’s already common in many processes, like financial reporting. Increasingly, companies are using AI to generate results quickly after the close of each quarter. But no business is simply going to paste the exact output into their regulatory statements. Instead, there will be several rounds of reviews to identify any errors.

The author of the research, Dr. Renée Richardson Gosline, often speaks of the value of adding speed bumps between AI output and delivery, what she refers to as “beneficial friction.”

In some cases, speed bumps are just the smarter strategy. One of our favorite recent examples of humans in the loop that we have done at Workato is a process that automatically identifies lost sales opportunities and drafts a “reactivation” email to the related contact.

The salesperson receives a notification in Slack with an AI-generated email for them to review, edit and approve. Once it is approved, the email is sent via Outreach on their behalf. To our sales team, this feels like magic. It is automating important but non-urgent work that doesn’t always get done.

In reality, it is an orchestration between the CRM (Salesforce), collaboration tool (Slack), sales engagement platform (Outreach) and an LLM (GPT4 by OpenAI). The speed bump—the human interaction—ensures that an email isn’t automatically deployed without valuable context only gained through years of experience and personal relationships with the clients.

The Importance Of Feedback

To get to that level of AI maturity, companies must start thinking about humans as a key ingredient in orchestrating their AI-enabled processes. Then, they can weave together people, systems, processes and AI to enhance productivity and efficiency.

As workers begin to interact with new AI applications, mistakes are bound to happen. That’s inevitable with new technology, especially one as powerful and complex as AI. But unlike in the past, when it could take years for software issues to get resolved, the immediate feedback that employees provide will be critical in helping to enhance the performance of the systems to reduce errors in the future. Feedback can be as simple as a thumbs-up or thumbs-down button. The feedback will make your LLMs smarter, leading to more intelligent processes over time.

This will be a game-changer for organizations. Improved models will encourage greater usage, increasing the amount of feedback the systems get and further enhancing performance. It’s this cycle that will help companies reap the real benefits of more intelligent enterprises. And it all starts with human-in-the-loop design today.

Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?

• Editorial Standards
• Reprints & Permissions

Salesforce is closed for new business in your area.