As the COO of a multinational business, I use AI nearly every day and strongly encourage my team to do so. While many tools are initially easy to use, wielding them effectively can be challenging, and staying current on their features and capabilities can sometimes seem like a job on its own. As a company, we have seen significant efficiencies and garnered valuable business insights that have allowed us to make better decisions faster, supporting the theory that AI has a greater ability to enhance/drive our efforts than to undermine them.
AI in Financial Operations
Google’s Gemini is well integrated within Google Workspace. We use Gemini AI within Google Sheets and Docs to analyze financial statements and other data sets, asking Gemini to identify patterns and outliers, create summaries and charts, and also modify parameters and run comparisons. This minimizes the time it takes us to spot trends or identify anomalies. It is a highly valuable assessment tool to evaluate, summarize, spot patterns, and recognize edge cases. We have been able to see which expenses are increasing at a faster-than-expected rate and are trending ahead of forecasts, thus providing us the opportunity to take action sooner. It is also very efficient to instruct AI to make an assumption and rerun the analysis or forecast. What takes a minute or two utilizing AI would normally take days and the involvement of other team members who need to juggle priorities. These insights have allowed us to make decisions about vendors and SaaS providers that will save us six figures annually. While it is very likely we would have arrived at the same conclusions without AI, those decisions would have been made months later, thereby costing us tens of thousands of dollars.
AI in Marketing
For greater automation, we use AI as a research and workflow management tool. We are trying out different services—such as Google Gems, N8N, and Gumloop—to create multi-agent workflows that do research, aggregate data from multiple sources, synthesize the data, and then present it in a format we define. Actions such as automatically queued emails can be added. We use this when identifying leads, learning more about the company they work for (size, revenue, year founded, recent news), and analyzing the person’s own social profile and history. We can attribute those new data points to our lead profile in our Sales CRM and then draft communication outreach or send an assessment to our sales team. We use similar workflows to round out the data required for us to generate robust marketing assets such as portfolios of client work and case studies, saving many hours of research, all of which can be created quickly and tailored to specific prospects. It is imperative to ground the data when necessary and attribute the externally acquired data to its source, often instructing the agent to use specific sources such as Bloomberg or Dun & Bradstreet. This richer content has seen a 40% increase in email open rates and a 250% increase in click-through rates, contributing to roughly 25% more calls booked.
AI in Recruiting
Our recruitment team uses a similar workflow process to learn about competitors in the marketplace. We pull company information from submitted resumes, gather company information using our AI research agents, and supplement the data set with information on open positions, company reviews, and publicly available salary and employee satisfaction data. We can combine this with the information applicants provide about the current state of the company and the reasons they are looking for a new position and then create targeted email campaigns to recruit employees directly from that competitor—and often reach employees with specific skills using LinkedIn Recruiter and other relevant social media platforms. Once again, most of these tasks could be done manually but would take dozens of hours and rarely make the top priority cut. Having these automation tools allows us to better understand the shifting market and be more proactive in identifying and hiring key personnel. We have also moved to a new ATS platform that has AI tools built in, allowing us to read summaries of a candidate’s complete application: resume, cover letter, and responses to questions and assignments. The platform also scores applicants based on our ranking, saving valuable time to prioritize top candidates and meet with them sooner. Additional tools, familiar across most video conference platforms, generate summaries of conversations in conducted interviews.
AI in Software Development
For our software development team, not only are they building applications with AI features, but they are also using many AI tools to significantly improve development velocity. Tools act as powerful accelerators, automating repetitive tasks, generating boilerplate code, and providing instant insights. Tools like GitHub Copilot exemplify the “pair programming” paradigm with AI. Trained on vast amounts of code, Copilot can suggest lines of code, entire functions, and even complex algorithms in real time. This significantly reduces the time engineers spend on boilerplate code, syntax recall, and common patterns. Beyond simple suggestions, AI can generate more substantial code segments, freeing up engineers to focus on higher-level architectural decisions and complex logic rather than tedious implementation details. AI can also assist in refactoring existing codebases, identifying areas for improvement, and suggesting optimized alternatives, leading to cleaner and more maintainable code. AI can generate test cases, analyze code coverage, and even predict potential failure points. This allows for more comprehensive testing and proactive bug detection. AI development that integrates human intelligence with machine learning to enhance accuracy and ensure the integrity of AI-powered products is key for stakeholders and developers alike. For software developers, Human-in-the-Loop (HITL) is necessary because, while it excels at processing vast amounts of data and identifying patterns, AI often lacks the nuanced understanding, contextual awareness, and ethical reasoning that humans possess.
Fundamentally, whether as an engineer or business stakeholder, we most often pay attention to the following AI tools and practices:
- Retrieval-Augmented Generation (RAG) for context-aware code generation
- Fine-tuning models on client-specific sources
- Chain-of-Thought reasoning for complex decisions
- Human-in-the-Loop validation for mission-critical applications
Retooling Skillsets
The ironic thing is that we used to applaud people for being specialists, and now the biggest skill from a training perspective is being a generalist who can bridge the gap between technology and business stakeholders. Skills needed to work with disparate AI tools will also be essential in interpersonal collaboration, such as clearly articulating perspective, needs, and context. A great AI prompt is also a great way to clearly express oneself to a colleague, client, or partner—higher yield and more efficient. But the true skill underlying all of this is problem-solving. Albert Einstein is famously quoted as saying something to the effect of, “If I had an hour to solve a problem, I would spend 55 minutes defining the problem and five minutes finding the solution.” AI has allowed us to spend more time ideating, defining the problems we wish to solve, and vastly expediting how to deliver viable solutions.
AI Risks
There are tremendous benefits, but not without risks. The most significant are relying on AI outputs without validating them and the security of confidential data. There are many vulnerabilities that can be exploited: untrusted data sources, hallucinations, and actions that can be taken on the user’s behalf that may not be appropriate. When we let AI run unsupervised, we dramatically increase the likelihood of negative outcomes and missed opportunities. If we maintain control of how and where we are deploying our AI tools, like a puppetmaster over a marionette, we gain more confidence, consistency, efficiency, and predictability of outcomes.
AI is driving headcount down—and will continue to—in most areas. It is no wonder that customer service representatives, translators, writers and authors, data entry clerks, and legal researchers are those whose jobs are most impacted by AI in 2025. We should absolutely expect and proactively plan for doing more with less. Sometimes, this will mean substituting three junior employees for one with more experience and skill. In other instances, we may be able to eliminate higher-paying jobs that no longer drive the necessary ROI.
The AI Wrap-Up
There is much AI hype, and we should be cautious not to rush into relying exclusively and independently on AI. It is, however, a tremendous tool that is quickly shaping many aspects of our lives. To stay current, protect our company’s interest, and fight to remain relevant, we must leverage AI as a powerful and essential tool in all aspects of our work.
