I think we all understand content slop by now – the low value content pumped out by AI (usually at scale).
Annoying, yes – but mostly harmless. One of the lacklustre side effects of AI advances.
However, there’s another type of AI related slop coming – what I’ll refer to as ‘process slop’.
It’s the side effect of people taking the ‘advice’ of AI when it comes to setting up their company processes and systems.
Business Trends
Perhaps you’re seeing more of this lately: a push within companies to self-solve internally, encouraging staff to self educate (often in their own time), and – not coincidentally – avoiding external agency costs as much as possible.
I’m all for self development, and I agree we should be continually striving to learn and grow our skills. And a big part of that currently is using AI tools. All good so far.
Domain Knowledge Problems
The problem comes however, when users don’t have a good grounding in the tools and platforms they use. They can’t tell whether the guidance the AI tool gives us is actually efficient or not. They don’t know what they don’t know…
A simple example to illustrate.
One of the first things I like to do when playing with a new AI tool is ask it this question:
“What are some common HubSpot workflow approaches for managing life cycle changes?”Based on the responses I can quickly tell me how reliable it is in understanding HubSpot and business processes.
Recently I played with Le Chat (it was flavour of the month a month or so ago).
You can read the full details of the guidance it provided in my post here, but the summary was: it provided a bunch of suggestions that sounded useful (perhaps even insightful), but were mostly bad advice.
Admittedly some were good (the first two for example in that post), but the others were busywork at best, and maintenance/technical debt at worst.
The suggestions included creating processes and workflows for things that were already controlled by simple settings, through to vague concepts that would send new users down rabbit holes of confusion (eg ‘build a partner nurture workflow based on joint marketing and co-selling activities…)
The Tip of the Iceberg
Concerningly, we’re starting to see the effects of this already appearing in prospect’s portals – in short: they’re becoming a mess.
Here’s the types of things I’m seeing:
- Unnecessary workflows (replicating existing simple settings)
- Complicated workflows (lots of complex triggers and multiple branching)
- Duplicated properties (created for storing interim data for criteria that wasn’t needed in the first place)
- Complicated and Unnecessary lists (attempting to segment these unnecessary properties)
- Attribution impacting changes (eg overwriting conversion properties)
- Complicated Sales pipelines (eg unnecessary additional deal stages, with confusing naming)
I can understand why users in companies go down this path – they want to save costs and time: why wait to work with an expensive HubSpot consultant tomorrow if they can get 80% of the way there using ChatGPT today? Or so the logic goes.
False Economies
But I fear that the cost savings may come back to bite, when they (or their replacement) realise in a year’s time that their HubSpot portal is out of control and they need to pay the ‘expensive consultants’ to come in and clean it up.
(Aside: I think there’s probably a lucrative ‘Fix up the mess AI made in your portal’ consulting opportunity here…)
Impact Isn’t Equal
Note though: The impact of the mess is related to the area of the business. For example, using AI tools to build your new website (assuming it’s just a marketing focussed website) is probably fine – it doesn’t matter about all the bloat and technical debt because when you refresh the site in a few years time you can just start over.
But when it’s internal business processes (eg sales pipelines, marketing nurtures, management reporting, customer support response procedures), and you’re building the platform that serves as a foundation, it’s really important to get it right. You need to avoid unnecessary complexity across processes and data architecture.
HubSpot of course has coverage across all these areas – websites through to CRM data structure through to management reports. So you need to be careful which parts you cut corners on.
Accelerate (Don’t Substitute)
To reiterate though – I’m not anti-AI. I’m definitely all for it. We use it all through our own business.
But we use it as an accelerator, not a substitute, for strong domain knowledge.
Next Steps
How then to address this?
There’s no easy answer, so I’ll just mention the high level approach, simple as it is:
The first step is to acknowledge that there’s a looming problem, and be careful about the processes you implement based solely on AI guidance. I’m going to assume that’s reasonably self-evident.
The next step, and most important, is to be clear about which areas of your business are crucial to get right:
- What are the foundations that need to be solid? Don’t skimp on those areas. Work with experts.
- But move to a Done With You model where possible (as opposed to Done For You).
- Aim to train your teams using experts – to equip them to implement (ie accelerate) with AI tools.
Then, organise quick, regular reviews with experts to validate the implementation (and course correct as required).
Notes:
- [1] Companies use so many tools these days, it’s rare that many staff are going to be experts on using them. And if they are, they won’t across all the tools, it will be a small subset.
- [2] There’s an old business maxim that you can tell a dishonest consultant because they add complexity to the situation (good consultants will simplify). This can potentially help when reviewing the guidance from an AI tool. Is it making thing simpler in your mind, or it is adding complexity? If the latter, then you perhaps you need to reconsider the approach.
- [3] Yes, I’m a consultant, and yes I can help you review your HubSpot portal (book time with me here if appropriate).
