In short, as per Part 1, you
- keep admitting to every mistake you are making and do something about it, then
- continue by looking for cost-effective opportunities for improvement and pursue them and finally
- never, ever, ever forget the timeless basics.
Today, we’ll continue by describing what you do when you identify, and admit to, one of the next two mistakes (mistakes 3 & 4) we chronicled in our two part introduction to our “dead company walking” (Part 1 and Part 2) series (where we helped your potential customers identify problems that signify you are a SaaS supplier they should be walking away from). (You can find Part 2 here.)
3) Shiny New Tech is More Important than Tried and True Methodology
Tried and true ALWAYS trumps shiny new tech in enterprise software. Especially when the total cost of ownership after factoring in license fees, maintenance, hardware & software updates, services, data feeds, integration fees, etc. usually push a solution into the realm of seven figures (and eight figures for large enterprises). Companies want a return, and shiny doesn’t generate a return. So focus on algorithms, processes, and approaches that are guaranteed to yield solutions. Do this by:
i) As THE REVELATOR would say, take an agent-based approach and focus in on what the solution should do
It’s supposed to be people-process-technology (or, in the doctor‘s words, talent-transformation-technology since you should first look for opportunities to improve your processes before investing in technology), not technology-forced workflow-prisoner! When you focus on the tech first and forget the process it is supposed to support and the people who have to use it, you’re never, ever, ever going to get it right. And believing that an over-hyped unproven technology will eventually show emergent behaviour (that it has zero chance of doing because of the underlying technology) and “evolve” to solve the problem is just stupid.
Real tech encodes real solutions identified by real human intelligence that have been implemented, repeatedly verified, tested in real world operating conditions, and understood to the point that the success can be repeated predictably. Sometimes it’s traditional AI, sometimes it’s RPA, sometimes it’s workflow, and sometimes it’s a simple calculation. It all depends on the problem and the people-process combo that can be applied to solve that problem. Tech is only transformative when it supports the business. You don’t lead with tech, you follow.
ii) Then identify the best tech options for each step
Once you have fully documented the problem(s) the people need to solve, the processes they can support, and the solutions that will work acceptably (maybe not perfectly, but it’s rare that a solution is perfect — plus, as we’ve repeatedly indicated, most enterprise users would cry tears of joys if they found something that just worked), identify all the potential tech options at your disposal, from open source to out-of-the-box from third parties to custom development.
Then, for each option, evaluate:
- it’s cost
- the relative return (for the customer and how that will translate into sales)
And identify the options with the best balance. You’ll be surprised at how often it’s not the new shiny. The best hotness is the old busted hotness.
iii) Select tried-and-true when all things are equal (and save)
Finally, when all things are equal, go for the most stable, tried-and-true, methodology — don’t do experimental new “AI” development if a sold optimization or analytics-based solution already exists.
4) Over Reliance on Third Party Tech is a Sustainable Business, Especially if its Gen-AI!
Reliance on third party tech, especially that which has not been proven to be on stable ground in the market, is not a good business plan. What do you do if the provider goes bankrupt, or, realizes they are on the path to bankruptcy and turns the tech off? If you have nothing else to pivot to, you go bankrupt too. Not a good scenario!
Even if you are building on tried-and-true stable tech (with a large install base), unless your business plan is to be acquired by the tech you’re building on and you actually have a chance of making that happen (you came from the provider, have good connections, etc.), your options are very, very limited if you don’t succeed.
You need to focus on a solution to a real problem that companies will pay to address first. And while it can be based 100% on third party tech to start, that shouldn’t be the game plan. The plan should be to acquire it, build something better in house, or at least find three or four options that will serve the same function until you can acquire or build appropriate tech. (The exception being a product maintained for you by a third party that is based on open source that will always be there for you to take ownership of and then assign to a new team.)
Once you have a solution, and know what tech you need, do the cost benefit analysis of:
- licensing someone else’s tech
- acquiring the third party tech
- building on, and contributing to, open source
- building your own
for each aspect of the solution, as well as the points at which one option becomes better.
In addition, even when you select a solution, always keep your eyes out for an alternative that has been demonstrated to be more reliable, efficient, or cost effective. When a problem has been newly identified, some companies will just throw anything at it to see what appears to work, without doing a proper analysis and designing a proper solution from the ground up. Eventually, some smart minds powered by Human Intelligence (HI!) will come up with a better solution. Once that becomes economical, you’ll want to switch to that if you don’t have an equivalent solution in house, while keeping back-up options open.
And definitely don’t (over) rely on third party Gen-AI — even the biggest companies can only afford to bleed for so long. There’s nothing Gen-AI can do that traditional tech can’t. It’s best uses are as an interface layer for (very) low TQ (Technology Quotient) folk who are being forced to use tech but just don’t get it (as a more natural language chatbot interface) or as a massive document store search and summarization solution (when traditional semantic / neural networks would just be overloaded). That’s it. And in all of these cases, it needs an underlying application that actually does the work and manages the data.
Stay tuned for Part 4!