Just in case you haven’t heard, the hot new buzzword (okay, buzz-phrase) is “generative AI.” It’s defined as “… algorithms (such as ChatGPT) that can be used to create new content, including audio, code, images, text, simulations, and videos.”
By now we’ve all been inundated with alarming accounts of generative AI’s societal consequences (including this, from me). And there’s a lot to be alarmed about, from the erosion of reality I wrote about in the referenced KJR to the potential obsolescence of Homo sapiens. To which, let me add this thought, in the interest of being first to ring the bell: What’s the big deal? Thirty or more years ago, pedagogues worried that the advent of cheap 10-key calculators would be the end of mathematics education as we know it. Now, there no longer seems to be much of a point to memorizing the multiplication tables.
The question: In the future, will we look back and wonder what all the fuss was about? Because really, isn’t generative AI’s impact just like that of the 10-key calculator but for verbal skills instead of arithmetic?
As a writer, consultant, and IT industry wiseguy I take such things personally. So to discover whether I’m already obsolete or not, I decided to put generative AI – OpenAI version – through its KJR paces by askingsome questions thatare central to KJR’s mission, starting with this: What are the most important factors in making IT organizations effective?
ChatGPT responded with a 600-word essay. Its core points:
- Leadership and Management
- Skilled Workforce
- Clear Goals and Objectives
- Effective Communication and Collaboration
- Agile and Adaptable Mindset
- Robust IT Governance
- Customer Focus
- Efficient Resource Management
It followed each of these headlines with a concise, explanatory paragraph.
For contrast, here’s the framework I’ve relied on for just about forever – Effective IT masters:
- Business integration
- Process maturity
- Technical architecture
- Human performance
My first reaction to ChatGPT’s answer was dismay. ChatGPT’s essay was, I had to concede, quite good.
But then I compared the two more closely. I’m delighted to report that every element of the ChatGPT framework fits into the KJR framework. Leadership and Management, and a skilled workforce, for example, are two sub-factors among a half-dozen that are subsumed by Human Performance.
The reverse isn’t true – the ChatGPT framework ignores, for example, the impact of a strong technical architecture, not to mention the critical role process maturity plays in the effectiveness of not only IT, but all organizations.
Ready for another? I asked, “What are the eight tasks of leadership?”
The answer (again, just the headlines):
- Setting Vision and Direction.
- Communicating and Influencing.
- Building and Managing Teams
- Making Decisions
- Planning and Organizing
- Managing Change
- Developing Others
- Ensuring Accountability and Results
KJR’s eight tasks:
- Setting direction
- Delegating
- Staffing
- Decision-making
- Motivation
- Managing team dynamics
- Engineering culture
- Communicating
Again, ChatGPT’s response was impressive – well-written, concise, and plausible. But digging in, it misses quite a bit. For example, the closest it gets to delegating is “developing others,” but that inverts purpose and fringe benefit.
As for communicating – arguably the single most important of the eight leadership tasks, ChatGPT entirely ignored listening and facilitating.
So I decided to give ChatGPT one more chance: “Is ‘internal customer’ a good idea?”
Its answer was, I have to admit, quite good – how the winning applicant for an open CIO slot might have responded. Again, summarizing:
Positive aspects of internal customer concept:
- Encourages collaboration
- Enhances service mindset
- Emphasizes employee satisfaction
Limitations of internal customer concept:
- May promote silos
- Can blur roles and responsibilities
- May not reflect organizational dynamics
Not bad. Personally, I think it’s too kind, but then I’m heavily invested in the end of the internal customer concept, so I can’t claim to be unbiased.
Bob’s last word: So I asked ChatGPT one more question: Should CIOs consider Bob Lewis a trusted advisor?
Its answer: “As an AI language model, I cannot make specific recommendations about individuals or their trustworthiness. Whether a Chief Information Officer (CIO) should consider Bob Lewis, or any other individual, as a trusted advisor would depend on various factors, including the CIO’s specific needs, goals, and context, as well as Bob Lewis’s qualifications, experience, and reputation.”
“Bob Lewis is a well-known author and consultant in the field of IT management and organizational effectiveness. He has written books on IT leadership and management, and has provided consulting services to organizations on IT strategy, governance, and operations. His insights and perspectives may be valuable for some CIOs seeking guidance in these areas.”
I’ll give ChatGPT points for diplomacy, and I guess I should be pleased that I appear on its radar screen at all.
Bob’s sales pitch: Please don’t think I’m encouraging you to shift your loyalties from KJR to ChatGPT. The latter’s responses are certainly plausible, but that isn’t the same thing as standing up to scrutiny (I nearly said “close scrutiny,” but if it’s scrutiny it’s close by definition).
Looking forward to a new weekly newsletter: KCR ‘Keep the Chat Running’ by ChatGPT.
(Of course, without prompting, what do you get?)
Well sure. You aren’t the one who’s brand is being threatened by a machine.
I have not been all that impressed by ChatGPT. It felt like the responses to question I posed it were what I would expect from an intern with decent language skills and time to Google answers. And pressed on a point it resorts to “As an AI language model I cannot …” you name it. It reminds me of a former coworker who would throw up reams of data but when asked to give a recommendation always deferred to others. He couldn’t understand why those of us who stuck our necks out were promoted ahead of him.
As my consultant I would want you to: A) Provide insights that I can’t just find with a web search, and B) based on evidence and your insights, have the courage to say “Do *this* instead of *that*.” (Even if in the end I decide to do *that* I would appreciate a forthright response.)
I think you job is safe. For now anyway.
BIG MISTAKE, Bob! In no time at all, this column will have been scanned and assimilated into the cube, leaving you without a counter-argument the next time someone asks ChatGTP, “Can you do a better job advising me than Bob Lewis?” Whoops!
However, I’m glad Sarah Huckabee Sanders did not use the AI to write a Republican response to the SOTU speech, as I did recently. Hers was MUCH more entertaining, and the ChatGTP version would not have inspired me to post this:
https://www.dailykos.com/stories/2023/2/9/2152132/-Republicans-just-cannot-understand-irony
ChatGPT sounds like a (micro?) manager instead of a leader….
The real issues that arise of the proliferation of Chat-bots won’t be due to lost employment opportunities. They will arise from the uncritical acceptance of fuzzy approximations they generate. Looking back the to times of batch reports printed on green bar paper, the saying was GIGO short for “Garbage In Gospel Out”
Or, as James Thurber wrote in “The Owl Who was God.”
Moral: You can fool too many of the people, too much of the time.
I don’t really mind generative AI’s GIGO potential. I worry more about the inverse GIGO effect: Gospel In, Garbage Out.
Re “internal customer”, ChatGPT didn’t mention the very real risk of the “That Way Lies Madness” scenario, or the various forces that can push an organization toward it.
If the IT function is serving “internal customers”, then the IT function itself is an “internal vendor”. Which can be easily replaced by an EXTERNAL vendor, i.e. let’s just outsource the whole thing.
It’s not as if IT is STRATEGIC, after all.This reveals the actual costs of the IT function more clearly than before: the cost of IT is the checks that you write to the external outsourced vendor.And if IT can be outsourced, so can lots of other departments/functions. And whatever residue of the organization remains, can now be divided up into Cost Centers, and we can now track minutely how much each Cost Center costs: it’s just the total of the checks that it writes to its outsourced vendors! Plus virtual checks that it writes to whatever “internal vendors” remain.
Everything’s a Cost Center! That guy in the cubicle is a cost center! That other guy in the cubicle next to him is a cost center! That woman in the office over there, the boss of the 2 guys in the cubicles, SHE’S a cost center! (And the 2 guys’ monthly income-and-expense statements roll up into HER monthly income-and-expense statement.)
Guy In Cubicle: “I need to cut costs in my cost center. Hey, Office Manager, I need to move to a smaller cubicle with fewer square feet. My monthly rent payment for my per-square-foot share of Common Area Maintenance is killing my income-and-expense statement!”
Real World: The fluorescent tube above Guy In Cubicle’s cubicle burns out. Fluorescent tubes are long and skinny. Half of it is above Guy In Cublcle’s cubicle. The other half is above Next To Him’s cubicle.
Guy in Cubicle: “I need to cut costs in my cost center. The overhead costs on paying Maintenance to replace the fluorescent tube will kill me. I don’t have any money to pay for it. YOU pay for it!”
Next To Him: “I don’t have any money either. YOU pay for it!”
Guy in Cubicle: “No, YOU pay for it!”
“YOU pay it!”… “You pay it!”… “YOU pay it!”… “You pay it!”
I once did computer consulting for an outsourcing firm, one of whose customers was an outfit I’ll call Gargantuan Corporation. Gargantuan Corp. had about 6000 employees, at about 600 retail stores (and kiosks) plus a Headquarters. Each retail store counted as a Cost Center — fair enough. Major departments at Headquarters, like Human Resources and Accounting, were Cost Centers — fair enough. IT was a cost center, with major divisions like Network Support and Macintosh Support rolled up into it — fair enough. But this organization with 6000 employees, 600 retail stores, and a headquarters, had 1800 cost centers in its tree-structured accounting hierarchy. Lower-level Cost Centers rolled up into higher-level Cost Centers. Every month, 800 income-and-expense statements were prepared, printed on paper, and physically distributed to 1800 Cost Centers. The tree-structured accounting hierarchy had 12 levels; by comparison, the Catholic Church, a much larger and much longer-lived organization, has only 7 levels — 8 if you count God Himself.
A lot of life inside of Gargantuan Corp. revolved around Cost Centers, and their 1800 monthly income-and-expense statements.
Yup, my 2 stories above, starring Guy In Cubicle, were actual real-world events that actually happened within Gargantuan Corp.
By the time I came on the scene, Gargantuan Corp. had already been bought out by a competitor that was a small fraction of the size of Gargantuan Corp. — but more profitable, which is how it could afford to buy out Gargantuan Corp. in the first place. The new ownership was NOT participating in all that Cost Center accounting, though preparations had been made so that the new ownership could be incorporated into it.
My consulting gig came to its natural end, and I fell out of contact with Gargantuan Corp.; the whole Cost Center thing was still grinding away, out of sheer inertia and commitment, but I suspected that its days were numbered. I could be wrong, though: Gargantuan Corp. wasn’t actually MERGED into its smaller owner, but was owned via a holding company. It is possible that the whole Cost Center accounting system, slightly updated, survives to this day.
Wow – that’s quite a story! Thanks for taking the time to share it.
Edited to add: yes, it’s 1800 income-and-expense statements, not 800.
“I’ll give ChatGPT points for diplomacy, and I guess I should be pleased that I appear on its radar screen at all.”
It could be much worse. You might actually be a ‘delusion’.
See: https://en.wikipedia.org/wiki/Hallucination_(artificial_intelligence)
Ignore this Comment, humans! It is just another example of anti-AI propaganda!
You compared ChatGPT to your work. The ChatGPT responses also sounded like any number of other columnists’ banal ‘insight’ over the years.
I guess its results are what we’d expect if we were somehow able to compute the verbal average of what’s been written about a subject. Anyway, thanks for the reassurance that I haven’t been excessively banal.