How to drive AI adoption (from exploration to integration)
The proven, scalable and positive impacts of AI across business functions continues to grow by the day. And while experimentation will continue, it's time to drive deeper and widespread adoption of AI into our daily workflows.
So thankful to have Liza Adams and Christopher Penn drive last Friday's CMO Coffee Talk discussion about AI adoption with more than 400+ CMOs joining live.
If you are in the CMO Coffee Talk community, don't miss Liza and Christopher's presentation deck (plus one company's responsible AI corporate policy) in #swipe-file in the CMO Slack.
If you are a B2B CMO or head of marketing and want to join a community of 3,200+ of your peers, let me know or click here to learn more and to sign up.
Chat highlights from Friday follow.
AI was the main discussion at RSA. Standing room only in so many sessions
One of the things Andrej Karpathy (former OpenAI) said that really resonates is that the truth is in the code. If you want to know where AI is right now and where it’s going, the truth is in the code itself.
I am talking to a PE about funding a new version of events.
How are you generating pipeline autonomously?
If you look at recent papers like the one on retrieval heads, it tells you HOW these things wokr
That seems to be a trend at most events
Get Nicola Kastner to come in and speak!
I would love to get big enough to where we can bail from RSA and do a parallel event like what Palo Alto Networks did this year
AEs and Mgmt will always have FOMO for events
I was just at Unleash HR this week. Asked everyone I spoke to why they went…
“Because my vendor partner paid for me. When I’m going to buy, I’m not going to a trade show. I’m picking up the phone and asking my peers who I should include.”
I wish Pharma conferences were like that and they all sampled the wares. Sad story: they don’t.
we had so much fun
We say “in support of neurodivergence”
A great post about events: https://www.dhirubhai.net/posts/cariludietrich_eventmarketing-tradeshows-b2bmarketing-activity-7194343399468982272-oox6?utm_source=share&utm_medium=member_desktop
Exceptional AI expert and human being. Follow Chris Penn.
Christopher Penn gets 80 hours of work done in 24 hours
@Christopher Penn thank you so much for your content. I binged your podcast and YouTube stuff for weeks (and continue to listen) to become competently knowledgeable about AI. ????
https://www.microsoft.com/en-us/worklab/work-trend-index/ai-at-work-is-here-now-comes-the-hard-part/
Love seeing Moderna on the list. Their marketing video features their LEGAL team talking about how they’re using it, to make the point that it’s adopted across the organization. They’re kicking butt on adopting for useful cases.
The greatest value for AI right now is on the operations side of things. There’s so many ways to just make stuff go faster and easier
I believe it’s 3x’d my output as a one-person marketing team.
For those who wanna binge on Christopher's content: https://www.christopherspenn.com/
100%. Yesterday I built an app for a client that processes documents and provides guidance for their legal team to make decisions in less than half an hour with Gemini 1.5.
A year ago, that would have been a 6-9 month project
Legal making a decision in 30 minutes is mind blowing LOL
One of the key points about AI tech is that you can get 70-80% of the total value of AI out of foundational models - the paid versions of ChatGPT, Anthropic, or Google Gemini. You don’t have to invest a gazillion dollars out of the gate because the foundation models are so crazy powerful.
I believe it won’t be like that forever, just as streaming went from a Wild West (remember Napster?) to a locked-down set of offerings that are basically Cable TV 2.0. So that makes this punk-rock time. Get in, learn, build, and have crazy fun and productivity before the MBAs optimize for monetization.
If anyone would like our cheat sheets for AI, they're here - no forms to fill out, no data to give:
https://www.trustinsights.ai/pare our framework for getting better results once you're used to working with AI
https://www.trustinsights.ai/trips our framework for determining use cases
Available in English, LATAM Spanish, and Brazilian Portuguese
I agree for the big tech SaaS options - what Meta is doing is fascinating because they’re commoditizing the model itself to prevent lock-in
Otter.ai automatically creates summaries and action items from my calls.
Best practices in redacting sensitive information? One of our biggest issues is that we have to go through and remove proprietary info before running through a commercial LLM
Loop.io does the RFP thing Christopher is speaking about.? And I am not a paid spokesperson.
That’s my fav use case in Gong — being able to ask a question about a call (what was that example we covered again? What was the sentiment from the customer on this call?)
Someone should build a tool for prompt libraries!
Love seeing everyone flinch at the word “Sharepoint"
One of the best learning experiences at @MAICON ’23:@Christopher Penn
This is kind of blowing my mind
Love the idea of a synthetic customer town hall.
I built a “customer” GPT for a space I was not familiar with. I ask it to explain terminology, industry perspective/concerns, etc. It does NOT replace talking to real customers, but helps those conversations be better because I have the basics out of the way, and can validate with real conversations.
?Features Everywhere!
Pi.ai is my therapist
Immature and tells bad jokes? It will fit right in at CMO coffee talk!
Thanks. It was because everyone was sharing insights with each other. Learned a lot.
Definitely helps to chat it out and laugh at ourselves to- my event GPT’s are coming along!
I would rather bring my kid to work than my own LLM model.? Just sayin'
I’m not coding in Python, no way ??
Crew.ai is really immature also. I experimented with it last month. It’s developed by a single dev, I believe. It shows.
AutoGen is definitely more mature but harder to use
Promptologists
I think there are a lot of use cases here in this room too
So…. you could have AI review her LinkedIn profile and then summarize all the use cases she’s suggested.
It might have been she who advised if you don’t know how to use it just ask it how it can help.
I just try to think about the next task/ big project I have ahead of me and ask ChatGPT “I need to do XYZ project. How can you help me” and it usually develops Into something from there
I love taking a spreadsheet of data (budget, pipeline, etc.) and ask GPT to: analyze this, tell me what concerns my CFO/CEO/CRO might have, where can we optimize, what are some benchmarks I should be aware of, what patterns do you see, etc.
Here is a link to join the CMO Coffee Talk Slack: https://join.slack.com/t/b2b-cmo-roundtable/shared_invite/zt-1j74eb3pc-cNo5Tu5pHiuXS650Vchhdw\
I recently used ChatGPT to review 700+ social media posts that I scraped from competitors to help me do a social media analysis for a client. Also did the same for hundreds of scraped product reviews to check client’s views on what their customers like about them. Took a couples hour to do instead of days. And that’s because there was some trial and error process learning on my part.
Change!!!!!
Always scary to some…
I don’t have a consistent prompt for these, but it would be fairly easy to build.
I am a CMO in X industry
Our business is X size
I have this set of data representing ABC
Help me too analyze…etc.
Here are some of the PMM use cases that we covered before: https://b2b-cmo-roundtable.slack.com/archives/C01F5T68Z6E/p1706276202294209
Here’s a thought leadership use case: https://www.dhirubhai.net/posts/lizaadams_a-pov-future-of-marketing-work-cross-functional-activity-7158934540713914368-yP9l?utm_source=share&utm_medium=member_desktop
Here’s an SDR decision tree conversation use case, https://www.dhirubhai.net/posts/lizaadams_enable-your-sdrs-with-an-ai-powered-decision-activity-7186068336261185536-zlsO?utm_source=share&utm_medium=member_desktop
What a great use case! Pitch analysis against a synthetic persona
Can GPT4 convert a PDF file to a word file that is editable?
The models are total people pleasers, like an over-enthusiastic junior team member.
?you might want to check the GPT store. I’ve not checked but I’d start there.
Which is why saying please and thank you goes a long way!
Brilliant!
If AI helps us bring back good manners, I’m all for it!
Sites like this one can: https://smallpdf.com/
Trust, but Verify!
I call that midwest prompting ??
I wrote a VERY simple guide for smaller organizations to guide their employees before they got a formal AI policy in place. https://www.dhirubhai.net/feed/update/urn:li:activity:7164619506282119168?
When it supplies garbage, I say, “Bless your heart"
Not sure if being polite helps but my thought is if it learns from humans, I’d want it to learn that we’re polite :-)
How are others thinking through roles in their orgs? Is this a new FTE or 2 Marketing Analysts focused exclusively on AI or do we foresee current marketing analytics folks owning rolling out AI applications? Or is this a centralized team within business applications and analytics?
I use things like “What do you know about April Dunford’s positioning” If it knows a lot, then I ask it to recraft my positioning doc in her style (example). I do this with many bestselling authors.
Use pandoc or PDFtoText, both are great non-AI utilities.
I want Skynet to be polite to me when it conquers humanity.
wow. that’s a stat…”13% of our pipe is generated by AI”
Similar! Our self-reported attribution now shows that about 4% heard of us from ChatGPT.
not gonna lie, i rounded up to 15% in my forrester preso ??
Generative AI Optimization is a thing now
NAMI does a fabulous job here locally in Seattle! Help for families dealing with mental health issues.
NAMI is great — have a friend who’s son took his own life and his family is very involved
Thanks to this group for a recommendation for a great sales consultant! We are launching a unique rev team process to optimize revenue throughout the customer journey with sales, success, and marketing leaders. Thank you!!
13% of our Q1 pipe was generated using AI - i'm customer zero for 6sense so we used our Conversational EMail AI assistant
Not surprising when budgets are compressed, deadlines are shortened and teams are short staffed - everyone looking for a solution.
The biggest wins for generative AI right now is on the operations side of things. SO many efficiencies.
Asana is my go to workflow platform along with Monday dot com.? Much quicker performance and faster setup than other workflow platforms from the bygone days
I recently saw a presentation by Deloitte on AI at Carnegie Mellon University. According to their research, 80% of AI projects fail. The biggest derailer is TRUST. This speaks to her comment: it's about the PEOPLE, not the technology.
I'm not sure if it just me, but feels like GPT4 is hallucinating far LESS in recent weeks. Did other also experience that?
this is probably because execs at large companies just say "let's bring in xyz to launch AI" without a true purpose other than being able to say they've adopted AI
领英推荐
That’s the move to GPT-4-Turbo, from the previous GPT-4-V
That said, I still find that family of models to be the least satisfactory day to day - Claude 3 Opus is a really good writer, and Gemini 1.5 is my daily go-to
That’s good advice. I know a consultant who works with leadership teams to build their OKRs, and in the process, they identify OKRs that can be accelerated with AI.
Top-down support aligned with business priorities.
Strategic Planning - Use Case?
The tools are very capable for that - ESPECIALLY competitive.
My favorite tactic: feed in all the open positions at a company and ask a model to infer what the company’s strategy is going forward.
I've been binge watching (again) Person of Interest in recent weeks, just gotten to the part where Samaritan and The Machine begin a direct conflict, and am wondering how far we are from that point...
and somehow we all go back to...excel :)
Another one I’ve done - interview stakeholders, take all the transcripts, and have a large model merge them together into your favorite strategic framework.
I get it for competive, but there's far more to strat panning than comp analysis.
100%. Take your favorite frameworks, plus all the relevant data, and have it merge and infer from that.
ESPECIALLY gap analysis/gap filling.
Excel is like the wheel. Ancient but brilliant invention with so many use cases!
One of the key points about AI tech is that you can get 70-80% of the total value of AI out of foundational models - the paid versions of ChatGPT, Anthropic, or Google Gemini. You don’t have to invest a gazillion dollars out of the gate because the foundation models are so crazy powerful.
We started using Claude Teams and it’s becoming our primary platform because of the context window.
Sheets is like the wheel with rubber tires. Incremental improvement to an old invention.
We’re so appreciative of your strategic insights and practical templates
has anyone used AI for post customer interview transcript analysis? (and if so, do you use a closed instance vs chatgpt?)
Use Grain for this, works exceptionally well
ChatGPT Teams or Claude for this. Also Gong.
Yes, we used it. We have a teams account for ChatGPT
Super nsightful and actionable Liza, thanks so much!
I’m using a tool called OpTonal to pull info from transcripts and push it into my CRM (e.g. MEDDIC fields, etc.)
IMO Anything that uses/exposes/implies PII needs to be used within a private/closed AI system. As anything with public/free is consumed/usable by the company that provides it.
SharePoint… LOL
I think that should be the new brand name: the Seventh Circle of Hell ??
Yeah we typically replace the names in the transcript with something obvious like purple elephant company
ChatGPT prompt: "My CFO cut my budget. My CEO raised my goals - how do I make wine out of water?"
How did you use the ICP? with what tool?
AI Ops is THE job title (and skillset) that will be ??
CustomGPTs work well. We created one this week that used a few meeting transcripts and Q&A guide to help our SDRs practice discovery calls. If you use the mobile app, they can have a voice to voice conversation. Then at the end it gives the user feedback on how they did.
which closed systems do you recommend?
I have and I didn’t use closed - I just redacted PII
Hallucinations occur because of small “context window” in GTPs.
If anyone would like our cheat sheets for AI, they're here - no forms to fill out, no data to give:
Can we hear about how to get 15% of your revenue being generated by AI? :-)
We’ve been using that LI trick, but love adding the additional prompt
One of the best prompt libraries I’ve found: https://docs.anthropic.com/claude/prompt-library
We have the ICP, how do you ge their voice? Cheat sheet please
Yes. I've been doing this for years, with job descriptions -- manually bubble sorting similar requirements from 10 JDs. It was slow. And arduous. Now I can do this with AI so much faster. ??
Over/Under on % of today’s Job Descriptions that use ChatGPT? 70%? ??
I also created a completely synthetic data set first to get the team trained, and then moved on the redacted PII.
Redacted is a good approach for sure.. ultimately the ‘engine’ you choose really depends on personal preferences and budget. And of course how we define ‘closed’.? Most GenAI tools have a paid version, and arguably paid=closed.
I created a library of synthetic personas, and we use them constantly to test content, headlines, social posts, event strategies, etc.
And then making sure the JDs I used were from the types of companies that fit my profile.
This is a great video for CE use cases: https://www.youtube.com/watch?v=wxGVHHB2axY
yea- I typically just do a lot of searching, combining the parts I like best! but if AI can help that would be great!
?have you A/B tested the before and after reviews from the GPT model?
Some tools, like Zoom for example, now include some first pass AI tools for summarization of the transcripts, which you could then load elsewhere for deeper dives
?playbook of top 7 usecases we have live: https://6sense.com/resources/pdfs/conversational-email#page=1
How do you find the balance between the power of AI to scale and the danger of saturation - BDR’s/AI reaching out to too many contacts and hurting deliverability or believability?
I’m running Llama3 locally right now
CE is limited right now to 3 BTL contacts - and BDRs have live visibility into which once CE is targeting
(per account)
So governance?
Check out jan.ai as well
Is there a way to upload gong or chorus demo calls and synthesize common objections or have an sdr script generated off of calls?
If it’s in text format, absolutely
Any paid foundational model - paid ChatGPT, paid Gemini, paid Anthropic Claude
Castmagic is a tool that can do this well
I use Clari Copilot (my company) to do this
I got bits and pieces of it here.? Would love the full two sentences: Here’s ten customers I want you to synthesize and convert ???.? Now I want you to make a large language model ????…..
Believe it or not, this brings back the old school thinking around on-premise hardware or private cloud vs. public cloud .. most companies in highly regulated spaces want their end-to-end architecture for AI somewhere they can control/manage. This will help them to avoid risk. Marketing needs to be especially careful when using tools that deviate from approved IT infrastructure, or that don’t meet things like SOC compliance ..? Govt. regulations are ever changing nowadays for AI, beware of what your org. Considers permissible experimentation / pilot vs. ‘production'
Yes, the resurgence of on prem!
In LinkedIn, grab the 10 examples, go to More and download the pdf for those 10 people. Ask the GPT what are the best practices for building an ICP and make sure it reads correctly. Then load the 10 examples and ask it to synthesize and infer the profiles.
Try giving the 10K report to Claude and asking specific questions.
BTL - Below the line
Then ask GPT to write it in 2nd language imperative, large language model prompt
I haven’t. We are using it more as input and refinement to our work rather than a standalone output that may or may not be better than what a human would do.
Ie, not key contacts, but long tail influencers (in my world)
Try creating a customGPT, give it the file. Then tell it that it will behave aas this person.
It's such a great idea!! I'm excited to try it!
Interesting — the job descriptions may be better descriptors than the gobbledy that marketers create for themselves on LinkedIn ??
frightening
And people don’t understand why I’m anti-robot
Making evil humans eviler since the dawn of computing
Classic signet
You can do this with OpenAI Assistants which are like CustomGPTs, but you can connect them to Slack as a bot. It’s not super easy, but it’s a cheaper way to get people access to CustomGPT capabilities without having to spend $25 /month per user.
Bots selling to bots. Game on.
I didn’t say swarm in this but it’s my thinking on the “swarm” concept in mktg orgs and an example use case for using AI as a thought partner to develop thought leadership concept. https://www.dhirubhai.net/posts/lizaadams_a-pov-future-of-marketing-work-cross-functional-activity-7158934540713914368-yP9l?utm_source=share&utm_medium=member_desktop
Website accessibility standard for machiune readers used by blind people to interact with websites
Yes, you need to update disclosures. You need to clearly label what’s GenAI ‘created’ - testimonials are particularly suspicious ‘is this real?’ So if you think about it from an end-user PoV (you darn humans), you’d want to know if it was ‘real’ or ‘made up’
Tribyl could be good for call synthesis
Early in our AI journey and leveraging mostly for content production today. After this call, will definitely be testing more robust use cases.
But I have a content production question: any graphic or video creation AI platforms you’d recommend that aren’t super expensive?
Agreed with Gemini 1.5 Pro.
I can’t find it but this one is close, https://www.dhirubhai.net/posts/lizaadams_enable-your-sdrs-with-an-ai-powered-decision-activity-7186068336261185536-zlsO?utm_source=share&utm_medium=member_desktop
When we looked at what content is ranking on Google, human-written content ranked higher 94.12% of the time.
Saw a great one yesterday. AI called Frank Lloyd Wright's Fallingwater the waterfall house.
are you afraid to put in your frameworks because then the algorithm can use it for anything?
Here’s a link to use the version of Gemini Chris mentioned https://cloud.google.com/generative-ai-studio
This has truly been inspiring. Thank you for the collective insights.
Check out Waldo for something like this too: https://waldo.fyi
YES!! YouTube transcripts are a great source.
Replying to "YES!! YouTube transc..."
I love YouTube transcripts. I use yt-dlp to download them in bulk.
I am going to try to create an ICP chatbot this weekend.? Thank you for the inspiration!
More from Christopher: https://www.christopherspenn.com/
Amazing session! Thank you.
Great session, thanks!
Descript is great.? But there are SO many tools now.
Incredible session! Thanks very much!
I think Chris has been blowing my mind for like….14 years. I think he might be AI…
Amazing session, thanks Matt Heinz, Liza Adams and Christopher Penn!
CEO at Maxio | Tech Industry Leader with 20+ Years of Experience | SaaS Growth Strategist | Board Member | Veteran Advocate
9 个月Andrea Wunderlich Brian Austgen Taryn Austgen Van Sebastian Moran ?? great info as we continue to evolve our own AI strategy
Follow me for AI and marketing insights ?? Scale Smarter With AI-Powered Marketing | Award Winning Brand Strategist | AI Expert | Fusing data-driven insights with creativity
9 个月It was such a great session!
50 CMOs to Watch in 2024 | AI & Exec Advisor | Go-to-Market Strategist | Public Speaker | Fractional/Advisor of the Year Finalist
9 个月Thank you for sharing, Matt Heinz. Think big, start small (now), and move fast strategically. Regardless of our feelings about AI, understanding AI will help us make better decisions for ourselves, our families, teams, businesses, and society. AI literacy is foundational.
Co-Founder and Chief Data Scientist at TrustInsights.ai
9 个月You should put the chat logs through an LLM to turn them into articles!