Want to use AI for your business? Philly-area experts lend their tips.

As a technology consultant, I’ve had a growing number of clients ask me how they can start using AI in their businesses. Generally, I recommend four options, which can be used independently or together.

For starters, if you’re like most businesses, you use either Microsoft Office or Google Workspace as your office platform. But you’re probably underusing the AI capabilities of Microsoft Copilot and Google Gemini to do tasks like creating presentations, revising emails, managing files, updating calendars, translating conversations, and analyzing spreadsheets.

Consider training your employees to better leverage these features.

Many of my clients are starting to connect their AI chatbot — Claude, ChatGPT, etc. — to their data. They are either applying the out-of-the-box integrations that come with these platforms or using Model Context Protocol (MCP), which can connect these platforms to most data models, databases, and files.

Of course, it’s important to establish safeguards and policies around which applications are approved and what data can be used. Once connected, your employees can start prompting for analysis and assistance using your company’s actual data.

Chop Dawg founder and CEO Joshua Davidson uses his AI chatbot to help manage the day-to-day operations of the business, and he uses it to help clients do the same. Philadelphia-based Chop Dawg builds custom web, mobile, and AI-powered applications, and helps business use AI to improve operations.

“We connect AI to our accounting, project management, calendar, email, and website, and every morning, I get a briefing on where the money is and what needs attention,” he said. “None of it is compiled by a person.”

Many popular software applications and platforms — from Salesforce to QuickBooks to Paychex — are rolling out new customer relationship management, accounting, and human resources AI features that can be used to increase productivity. It’s important to reach out to your software vendors to better understand the ways you can use these new tools.

And then you can build your own AI application.

The key is defining an actual business problem that AI can address or solve, says Rick Tashman, managing partner of Tashman Tech in Drexel Hill. His company consults businesses on how to use AI to improve workflows and access company information.

“The biggest factor is having very clear intent and a very clear expected outcome,” Tashman said. “Projects often get into trouble when AI becomes a solution looking for a problem.”

Tashman recommends thinking backward: “Start with the problem and ask yourself what takes too long? What frustrates customers or employees? Where are people repeatedly searching for information? Then determine whether AI is an appropriate part of the solution.”

I’m seeing a growing number of use cases in the business world.

An AI application can, for example, listen in on conversations, transcribe what it hears, and automatically create quotes or proposals. Or, after being connected to multiple internal data sources, it can generate a weekly “CEO Report” of activities in your company. Or it can access your repository of manuals, technical documents, spec sheets, and other knowledge material about your products so employees and customers can prompt for answers, generate a service ticket, or create step-by-step instructions for fixing a problem.

After defining a problem, it’s important to address management’s commitment and the skills of the employees involved, said Brian Dainis, CEO of Philadelphia-based Curotec. His company builds AI-powered applications, automates business workflows, and modernizes existing software for midsized companies.

Dainis recommends establishing clear deliverables and making sure your company’s data is usable and built around a secure architecture.

“The biggest factors are: being aligned on expectations, staffing the team correctly, how bought-in the client is internally, how involved the customer is with our team, and whether the client has experience buying services from a technology firm,” he said.

“Technical feasibility is rarely the risk,” Dainis said. “The biggest risks are always the human factors.”

The cost of an AI project is always a major consideration.

Larger companies complain of the spiraling cost of their AI applications, according to media reports. These applications can eat up countless tokens — basic units of data, such as words, parts of words, syllables, or punctuation, that an AI model uses to read, process, and generate text — each time employees prompt their systems.

Most small businesses don’t have that kind of volume and haven’t seen these high costs … at least not yet. But, in my experience, the cost is not insignificant.

Up front, a 10-50 employee company can expect to pay thousands annually in subscription fees, $10,000 to $50,000 for a firm or developer to implement a solution, and then ongoing costs for maintenance and support.

Large modernization projects can take multiple years, Dainis said.

Davidson recommends an annual upkeep allowance of roughly 10%–15% of the build cost.

Tashman advised budgeting for increased usage, hosting, storage, and data-processing expenses.

“There are also ongoing costs that get overlooked — if the application runs through services such as Microsoft Azure or AWS, there may be charges based on computing, AI usage, storage, and the amount of data being processed,” Tashman said. “Businesses should evaluate the total cost over time, not simply what it costs to get started.”

Going forward, AI applications will likely be more tailored to individual businesses. Davidson expects these applications to take a more active role in identifying problems and preparing actions, while retaining an approval step. The agents are coming.

“The apps that win will act, not just display,” he said. “Today most software shows you information and waits. The next generation reads your business, flags what needs attention, drafts the follow-up, and asks you to approve it.”

My advice: Choose one process, establish its current cost or time requirement, and measure whether the application improves it.