In part one, we broke down the core differences between Copilot and AI agents in Dynamics 365: Copilot assists you in the flow of work, while agents act on your behalf to execute outcomes. Now it’s time to get practical, with real-world use cases for sales and marketing and a simple framework to help you decide which approach (or combination) is right for your team.
Part Two: Use cases and how to choose the right fit
It’s rarely “either/or”
Most organizations don’t end up picking Copilot or agents; they land somewhere in between, layering agents on top of Copilot-assisted workflows as their data, processes, and confidence mature. Think of it less like choosing a tool and more like sequencing a roadmap: Copilot gets your team comfortable working alongside AI, and agents take over the repeatable, high-volume work once you have the rules and data in place to trust it.
Use case: Inbound lead response
Copilot approach: A rep gets a new lead notification, opens the record, and asks Copilot to summarize the prospect’s firmographics and any prior engagement. Copilot drafts a personalized first-touch email; the rep reviews, tweaks the tone, and sends.
Agent approach: A Sales Qualification Agent monitors every inbound lead 24/7. It enriches the record, scores the lead against your ICP, and either books a meeting directly on the BDM’s calendar or sends a qualifying message via chat or email, no rep involvement until the lead is sales-ready.
Why it matters: Copilot speeds up the rep’s manual work; the agent removes the manual step entirely for high-volume, predictable scenarios, freeing reps to focus on warm conversations instead of triage.
Edge: Agent
Use case: Customer journey orchestration in Customer Insights
Copilot approach: This is a great example of Copilot doing the heavy lifting on setup. With Query Assist, a marketer can describe a segment in plain language, something like “contacts who attended our spring event and clicked an email in the last 30 days”, and Copilot suggests the segment structure for review, no filter logic required. Pair that with the Journey Creation Agent (currently in preview), where a marketer describes a campaign goal conversationally (“send a re-engagement email, then follow up with an SMS after three days if no email click”) and the agent scaffolds the full journey: steps, timing, and branching logic.
Agent approach: Conversational Journeys takes it a step further by combining Customer Insights – Journeys, Dynamics 365 Contact Center, and a Copilot Studio agent to deliver two-way AI conversations via voice or SMS, triggered automatically by your marketing logic. Think renewal reminders, expiring offers, or appointment follow-ups: the journey identifies who to contact and when (respecting consent and quiet times), the agent handles the conversation in natural language, and it can escalate to a human whenever the conversation calls for it.
Why it matters: Query Assist and the Journey Creation Agent dramatically cut the time it takes to go from campaign idea to live journey, often from hours to minutes. Conversational Journeys goes further, replacing what used to require a human conversation entirely. It’s worth noting these capabilities have real prerequisites (matching Dataverse environments, Contact Center licensing, Copilot Studio for the AI agent layer), so this is a case where the “governance and data readiness” question below matters before you commit.
Edge: Copilot (for now, as it’s not dependent on Contact Center)
Use case: Customer service case handling
Copilot approach: An agent (the human kind) opens a case. Copilot summarizes the customer’s history, suggests relevant knowledge articles, and drafts a response the rep can edit before sending.
Agent approach: A service triage agent classifies incoming cases the moment they arrive, gathers missing details directly from the customer, and either resolves simple issues fully via chat or routes complex ones to the right team with full context attached.
Why it matters: Copilot improves quality and speed for the agents handling cases. The triage agent reduces the volume of cases that need a human at all.
Edge: Tie.
Use case: Marketing workflow automation
Copilot approach: Content Ideas and Content Rewrite Copilots help a marketer draft email copy faster, and a Smart Recommendation Copilot surfaces the most relevant image from the asset library as the email is being built, no manual searching required.
Agent approach: Agents built in Copilot Studio can take over the operational work that happens between campaigns. Instead of an analyst periodically pulling performance reports and reviewing them manually (often after the window to act has passed), an agent monitors content metrics continuously, flags underperformance against benchmarks, and proactively surfaces specific recommendations. The same pattern applies to campaign briefing: rather than a strategist spending hours assembling a brief from scratch, an agent pulls historical performance and audience insights and delivers a structured starting point before the team opens a blank document.
Why it matters: Copilot makes the in-the-moment creative work faster. Agents reclaim time spent on the recurring, low-judgment legwork around campaigns, so your team’s strategic capacity goes toward growth instead of coordination.
Edge: Copilot (for now). While the current features are still a work in progress, extensive Copilot Studio usage may be a barrier for certain complex tasks. (More to come from Coffee + Dunn on this!)
Use case: Renewals and expansion
Copilot approach: An account manager preparing for a renewal conversation asks Copilot to summarize usage trends and flag any risk signals from recent support interactions.
Agent approach: A renewal agent continuously tracks contract dates, usage patterns, and risk indicators across accounts, then proactively sends personalized outreach timed to maximize the chance of renewal or upsell, well before a human would have noticed the signal.
Why it matters: This is a good example of where agents shine: outcomes that depend on constant monitoring across many accounts at once, which isn’t realistic for a human to do manually at scale.
Edge: Agents
A simple framework for choosing
As you evaluate, ask yourself:
1. How often does it happen, and how repeatable is it? High-volume, rules-based processes (lead routing, case triage, renewal monitoring) are strong agent candidates. Lower-volume or highly variable work is often better served by Copilot, where human judgment stays in the loop
2. How clean and connected is your data? Agents act autonomously, so they need reliable data across systems to make good decisions. If your CRM data has gaps or your systems aren’t well integrated, start with Copilot while you shore up the foundation.
3. What’s your risk tolerance for autonomous action? Sending a personalized email is lower risk than approving a discount or closing a case without review. Map your use cases by risk level, and reserve full autonomy for the lower-risk end until you’ve built trust in the agent’s outputs.
4. Do you have the governance to support it? Agents need guardrails: approval thresholds, monitoring, versioning, and a plan for continuous improvement. If that infrastructure isn’t in place yet, Copilot is the lower-lift starting point.
5. Where is the bottleneck, exactly? If your team is too slow at drafting and summarizing, Copilot solves that. If the bottleneck is throughput, response time, or coverage outside business hours, an agent is built for that problem.
Bringing it together
Copilot and AI agents aren’t competing strategies; they’re tools that solve different problems at different stages of maturity. Most teams will use both: Copilot to support the humans doing nuanced, judgment-heavy work, and agents to handle the high-volume, well-defined processes that don’t need a person in the loop every time.
The right starting point isn’t the flashiest AI feature. It’s the use case where your data, governance, and risk tolerance line up with what the technology actually requires to succeed.
Not sure where to start? Our team can help you map your processes, assess your data readiness, and build a roadmap for Copilot and agents that fits where your organization is today. Let’s talk.
Frequently Asked Questions
Do I need AI agents to use Copilot in Dynamics 365?
No. Copilot and agents can be used independently. Many organizations start with Copilot before introducing autonomous agents.
When should I choose an AI agent instead of Copilot?
AI agents are generally best for high-volume, repeatable processes such as lead qualification, case triage, and renewal monitoring. Copilot is often better for work that requires human judgment.
Can organizations use both Copilot and AI agents?
Yes. Most organizations will ultimately use both—Copilot for human-assisted productivity and agents for autonomous execution.
What are the prerequisites for Conversational Journeys?
Conversational Journeys require Customer Insights – Journeys, Dynamics 365 Contact Center, Copilot Studio agents, and aligned Dataverse environments.
How do AI agents differ from traditional Power Automate workflows?
Power Automate follows predefined logic paths and executes specific actions when triggered. AI agents can reason across multiple inputs, determine the next best action, interact conversationally with users, and adapt based on context rather than strictly following a static workflow. In many organizations, agents and workflows work together, with agents making decisions and Power Automate handling execution.
Can AI agents work across multiple Dynamics 365 applications?
Yes. One of the advantages of the Dynamics 365 and Dataverse ecosystem is that agents can leverage information across sales, marketing, customer service, contact center, and other connected business processes. The value often increases as more relevant data sources become available.
