When a business owner asks Total Technology Resources how their company can start using AI, there’s usually a bottleneck behind the question and a hope that AI is a modern way to fix it. Before we talk about tools, we talk about the problem.
“The first thing to identify is what they need the AI strategy to actually complete or to actually fix,” says Jim Smith, Managing Partner at Total Technology Resources. “To just deploy an AI strategy out of nowhere doesn’t necessarily achieve anything.”
This post is for business owners and leadership teams deciding whether a specific task is a good candidate for AI. If you’re not sure your processes are ready for AI at all, start with Don’t Automate a Broken Process.
What should you figure out before choosing an AI tool?
What the AI is supposed to fix. Jim starts by asking which current process isn’t efficient, or needs too much human involvement, and what you’d want AI to do about it. His starting questions:
- What daily tasks do you or your employees do that are repetitive?
- How much of your data entry is repetitive?
- What throughout the day feels very manual and could be automated?
Then put numbers on the answer. For example, if the task is processing invoices:
- How many invoices arrive each week?
- How many minutes does each one take to process?
- Where do the current errors come from?
- What is a 60% reduction in handling time worth to the business?
Which tasks are good candidates for AI?
“If you have a task that is being done by a person that’s very repetitive, AI would be a good solution for that,” says Jim. The strongest candidates:
- Moving information between systems: “If it’s asked just to copy or enter or pull out information from one aspect or one system to another, that’s when it tends to operate more efficiently.” AI can also bridge systems that don’t talk to each other, “with maybe a middleware or third party system,” without custom code. Jim’s examples are entry processes like order entry and customer entry.
- Reviewing numbers quickly: “Anytime there’s hard numbers or information that could take a while for a human to review and come up with an output, AI is great for those solutions.”
- Reading and summarizing text: AI can read emails and documents such as insurance paperwork and summarize them quickly. Employees are often already doing this on their own, which is why the account they use matters.
How we decide whether an AI project is worth pursuing
When a client asks whether AI could help them, Jim starts with what they’re trying to accomplish. “If they’re trying to accomplish something that removes some manual data entry or maybe a repetitive task, I give it some more thought. If they’re trying to actually replace somebody or replace a really complex system or complex software application, that’s where I take a break and step back to wonder what they’re ultimately trying to achieve.”
When is something other than AI the better fix?
When the process itself is the problem. Jim’s order of operations: “Identify the process itself, where is it lacking, what aspect of it is failing, and then identify maybe there’s software out there that can fix it, maybe there’s some sort of workflow change that could change it as well before just throwing AI at it.” In practice, that means checking:
- Software that already does it: Teams frequently pay for platform capabilities inside their existing CRM, ERP, or ITSM tools that were never configured. A week of process optimization inside a system you already licensed often removes more friction than a six-month custom build.
- A standard integration: “AI can create some bridges between systems without having to manually code it or manually connect them. AI can do that with maybe a middleware or third party system…” Standard connectors like APIs, an iPaaS layer, or scheduled data syncs solve that problem permanently and predictably.
- A workflow change: Removing a redundant approval step or changing an intake form can cut days out of an operational cycle at zero licensing cost.
Four questions to ask before approving a project
Before committing budget to a new AI build, run the proposed process through this evaluation sequence:
- Is the process mapped and fixed? If not, start there. AI on a broken process makes the flaw bigger
- How often does it actually run? For example, a task that takes four minutes and repeats 500 times a month uses about 33 hours of staff time, about $1,300 at a $40 loaded hourly rate. That math tells you immediately whether a large custom software build can pay for itself.
- What happens when the output is wrong? Jim’s rule of thumb, “Where AI mistakes can become serious is if there’s ever an opportunity for AI to create information… or any time AI is summarizing some information for a case and adding its own creation. That’s where we see AI really start to break down.” A person needs to check the results.
- How long will it take to build? Jim weighs this before taking on any AI project: “If I feel like it’s supposed to take really long, a lot of time, a lot of hours invested, that may not be good for both us or the client.”
Red flags that mean a process isn’t ready
Jim steps back from an AI project when:
- The goal is removing necessary human judgment: AI can reduce the work, but, “it’s not gonna completely remove the human in the loop for the process itself. There’s still a need for human interaction, just slightly less.”
- The process is still broken: “When you introduce AI to an already broken or flawed process, it just magnifies the flaw in the process itself. AI is not going to fix that process, it’s just going to add a more complex aspect to it.”
- There is no underlying system of record: “If they’re doing it themselves, are they using a service, maybe a free service where they’re inputting too much information to it?” Fix the foundational data layer and security first.
- Nobody owns the outcome: “The number one hint that they would have it as under control is they have someone in the company responsible for AI usage.”
Total Technology Resources three rules for using AI well
- Use it securely. Don’t let company or client data leak out to an AI chatbot or an AI company.
- Have one person manage it, “not just being managed by a conglomerate of employees.”
- Don’t rush to it because it’s a shiny new object. “You wanna make sure it’s the right solution first.”
“A pragmatic technology posture actually makes your eventual AI projects much easier,” says Jim. “When your processes are mapped, your data has clear ownership, and your core software trades information cleanly, an AI model finally has something solid to stand on.”
Before committing to a complex automation build, make sure your data layer and operational processes are ready. Start a conversation with the team at Total Technology Resources, and our IT services near Center City will help you evaluate your situation to see what solution the numbers actually support.
About the Author: Jim Smith
Jim Smith is a Managing Partner and Co-Founder of Total Technology Resources. With over two decades of experience in IT and systems management, Jim helps regional businesses navigate technology decisions, infrastructure design, and digital security. As a recognized leader in the managed services space, he works closely with client leadership to mitigate risk and optimize operational efficiency. der in the managed services space, he works closely with client leadership to mitigate risk and optimize operational efficiency.
