How Managers Can Harness the Power of AI to Grow Their Teams
Eliott Ardisson
Founder & CEO - Basalt Studio
A practical guide for managers at founder-led SMBs on using AI agents to cut administrative overhead, improve team visibility, and free people for higher-value work.
Key Takeaways
- AI agents are most valuable to managers not as a replacement for headcount, but as a way to absorb the administrative load that currently prevents focused work.
- The highest-ROI starting points are repetitive coordination tasks: scheduling, status tracking, reporting, and basic triage, not complex judgment calls.
- Successful adoption depends as much on change management as on technical implementation. Teams that understand why the change is happening adapt faster.
- McKinsey and Gartner research consistently points to 20–40% productivity gains when automation targets high-frequency, low-complexity workflows.
- AI doesn’t scale your team headcount. It scales your existing team’s capacity, which is a meaningfully different thing.
Running a 15-person team today often feels like managing 30. Not because the team is underperforming, but because a significant chunk of every week disappears into coordination overhead: chasing status updates, reformatting reports, scheduling and rescheduling, answering the same onboarding questions, and manually pushing information between tools that don’t talk to each other.
AI agents don’t solve all of that overnight. But deployed with clear intent, they can systematically recover that lost time and redirect it toward work that actually moves the business forward. This post is a practical look at how managers, particularly those in founder-led businesses, can do that without needing an internal engineering team or a six-month rollout.
What “AI for Team Management” Actually Means
It’s worth being precise about terminology before diving into tactics.
AI automation covers tools that follow defined rules to trigger actions: if a form is submitted, create a task; if a deadline passes, send a reminder. Useful, but brittle. These break when conditions change.
AI agents are a step further. They can interpret context, handle exceptions, and make conditional decisions based on instructions rather than rigid rules. An agent reviewing incoming requests can triage, categorize, draft a response, escalate to a human, or do all four depending on what the request actually contains.
For team management, the practical difference is this: automation handles the predictable. Agents handle the predictable and the slightly unpredictable, which is where most real workflow friction actually lives.
When this post refers to “AI for team management,” it means deploying agents, not just connecting tools with Zapier-style if-then logic.
Where Managers Are Actually Losing Time
Before picking tools or implementation approaches, it helps to look honestly at where management time goes. Most managers in SMBs report spending meaningful portions of their week on tasks that require their presence but not really their judgment:
- Pulling project status from five different places and summarizing it for a leadership update
- Rescheduling meetings when someone’s availability changes
- Writing the same onboarding checklist email for the fourth new hire this quarter
- Tracking whether follow-up tasks from last week’s meeting actually happened
- Answering questions that live in a document no one can find
None of these require a skilled manager. All of them cost skilled managers hours every week.
McKinsey research has consistently found that knowledge workers, including managers, spend a significant share of their working hours on tasks that could be partially or fully automated with available technology. The gap between what’s technically automatable and what organizations have actually automated remains large, particularly in smaller businesses that lack dedicated operations or IT resources.
The goal isn’t to automate management. It’s to automate the operational maintenance work that currently crowds out actual management.
High-Impact Starting Points for AI Implementation
Meeting and Calendar Coordination
Scheduling and rescheduling across a distributed team is time-consuming and cognitively annoying in equal measure. An AI agent connected to your calendar and your team’s availability can handle meeting booking, send prep materials pulled from the relevant project thread, and generate a brief summary with action items afterward.
This isn’t glamorous, but it’s immediate. Managers who implement this first often report recovering several hours a week within the first two weeks.
Status Reporting and Progress Tracking
Most weekly status updates are a manual aggregation exercise. Someone checks the project management tool, pulls numbers from a spreadsheet, and writes a paragraph. An AI agent can do the same thing on a schedule, pulling from connected sources and generating a draft summary that a manager reviews in two minutes rather than writes in twenty.
The key is connecting the agent to where work actually lives: your project management tool, your CRM, your shared documents. Without those connections, the agent can’t surface anything meaningful.
Onboarding and Internal FAQ Handling
Recruitment agencies, accounting practices, and professional services firms all have the same problem: new team members ask the same questions repeatedly, and the answers exist somewhere in a document that nobody can locate. An AI agent trained on your internal documentation can handle the majority of these queries, escalating only when the question is genuinely novel.
Task Triage and Routing
In teams that receive inbound work through email, shared inboxes, or Slack channels, triage is a constant low-grade tax on whoever is most responsive. An agent can categorize incoming work, assign it based on current team capacity or predefined rules, and flag anything that needs immediate human attention.
What Real Implementation Looks Like
There’s a tendency in AI content to make implementation sound either trivially easy or impossibly technical. The truth sits somewhere more mundane.
A meaningful AI implementation for a 15–50 person SMB typically involves:
-
A workflow audit: mapping where time actually goes, not where you think it goes. This is usually a one-to-two week exercise involving the manager and the people doing the work.
-
Prioritization: identifying the two or three workflows with the highest frequency, clearest definition, and most straightforward data connections. These become the first build targets.
-
Build and test: connecting the relevant tools, configuring the agent behavior, and running it in parallel with the manual process for a short period to catch edge cases before fully handing off.
-
Team training: making sure the people whose workflows have changed understand what the agent does, when it will get it wrong, and how to flag issues.
-
Ongoing refinement: checking performance periodically and adjusting when the agent’s behavior drifts from what’s needed.
In our work helping founder-led teams deploy their first AI agents, the most common breakdown isn’t technical. It’s that the workflow being automated wasn’t clearly defined to begin with, so there’s nothing concrete to hand off to the agent. The audit phase fixes this, but only if it’s taken seriously rather than treated as a formality.
The tools Basalt Studio typically uses for this kind of implementation include n8n for workflow orchestration, Claude API via the Anthropic SDK for language tasks, and TypeScript or Next.js where custom interfaces are needed. The specific stack matters less than having implementation partners who understand the operational context of the business.
Change Management: The Part Most Implementation Guides Skip
Technical implementation is the easier half of AI adoption. The harder half is getting a team to actually use the new system and trust it enough to let it handle work that previously required a human.
A few things that reliably help:
Name what’s changing and why. Telling someone their reporting process is being automated without explaining what they’ll do with the recovered time creates anxiety. Telling them specifically that the time currently spent on manual status updates will go toward client-facing work or project planning makes the change feel concrete and reasonable.
Start with things people actively dislike. If your team finds expense reporting tedious, automate that first. Early wins on genuinely unpopular tasks build more trust than optimizing a process people feel neutral about.
Involve the team in problem identification. The people doing the work almost always have better insight into friction points than managers observing from a distance. A brief structured session asking “what’s the most annoying thing you do every week that you’d happily hand off?” will surface better automation targets than any workflow diagram.
Accept that the agent will make mistakes. No AI agent is perfect in its first few weeks of operation. Setting this expectation upfront, and building in a clear correction mechanism, prevents early errors from becoming a broader rejection of the technology.
Gartner has noted that change management is consistently underinvested in enterprise AI deployments, with organizations typically spending three to five times more on technical implementation than on adoption planning. The pattern in SMBs is similar, with smaller scale but faster organizational impact.
Definitions: Key Terms for Manager Context
AI Agent: A software system that uses a language model to interpret instructions, process inputs, and take actions or produce outputs. Unlike a simple chatbot, an agent can execute multi-step tasks and make conditional decisions.
Workflow Automation: Connecting discrete steps in a process so they trigger each other automatically, typically without AI judgment, based on predefined rules.
Agentic Workflow: A more complex automation where an AI model interprets inputs, selects from multiple possible actions, and handles exceptions without requiring rigid rule definitions for every case.
LLM (Large Language Model): The underlying AI model that powers agents’ ability to understand and generate language. Claude (Anthropic) is one example used in production deployments.
Orchestration Layer: The infrastructure that manages how agents receive inputs, execute tasks, and pass outputs to other systems. Tools like n8n serve this function.
Common Implementation Pitfalls
Even well-intentioned AI rollouts go sideways in predictable ways. The ones worth watching for:
Automating a broken process: If a workflow is poorly designed, automating it makes it faster and worse simultaneously. Fix the process first, then automate it.
Underestimating data quality requirements: Agents pulling information from messy, inconsistently maintained systems will produce unreliable outputs. The data sources feeding your agents need to be reasonably clean before you can trust the agent’s work.
Over-building in phase one: Trying to automate everything at once typically results in a complex, fragile system that’s hard to debug and harder to hand off. Narrow scope on initial builds, prove the value, then expand.
Skipping the feedback loop: Without a mechanism for team members to report when an agent gets something wrong, errors compound silently. Build correction pathways from the start.
Treating implementation as one-time: Agent behavior needs periodic review. Workflows change, team composition changes, business priorities shift. An agent configured for last year’s process may be actively unhelpful by mid-year.
What to Measure After Deployment
Establish baselines before you deploy anything. Without them, you can’t make a credible case that the investment was worth it, and you can’t identify whether performance is improving.
Useful metrics to track:
- Hours per week spent on targeted administrative tasks (before and after)
- Time-to-completion for specific workflows (e.g., how long from project status check to summary report)
- Error or rework rate on automated outputs
- Team satisfaction with process quality (a simple periodic pulse check works here)
- Escalation rate: what percentage of tasks the agent handles end up requiring human intervention
The last metric is particularly useful. An escalation rate that’s very high suggests the agent’s scope is too broad or the workflow definition needs tightening. One that’s very low might mean the agent is handling things that genuinely needed human judgment.
Realistic improvement timelines, based on directional research from firms like Deloitte and Forrester, suggest that well-scoped automation projects show measurable time savings within the first 30 to 60 days, with productivity gains in the 20–40% range for targeted workflows over the first six months.
Building Toward a More Sustainable Team Operation
AI agents don’t solve management. They solve a specific category of problems that currently prevent managers from doing management well.
The businesses that get the most out of this aren’t the ones that implement the most agents. They’re the ones that implement deliberately, starting with genuine pain points, getting buy-in from the people affected, and building a culture of iterative improvement rather than one-time deployment.
For founder-led SMBs in particular, the opportunity is significant. These businesses tend to be agile enough to implement quickly but resource-constrained enough that every hour of management time matters. Recovering even five hours per week per manager, applied to coaching, strategy, or client relationships, compounds meaningfully over a year.
If you’re working through where to start with AI in your team operations, Basalt offers an AI strategy call where we map your current workflows against realistic automation opportunities. No pitch deck, no fake ROI projections: just a concrete conversation about what’s worth building for your specific situation.