Free AB-210 practice test: 30 questions with answers

Copilot in Dynamics 365 Sales summarizes account and opportunity activity so representatives can prepare quickly. Manual exports, printed notes and monthly reports are slower and less current.

Copilot can draft emails using record context such as activity and stage, which saves time and keeps content relevant. Blank templates, printed letters and voicemail scripts do not use the record context.

AI-generated summaries should be checked against source records before they are shared, because they can omit or misstate details. Trusting outputs blindly, deleting sources and involving customers in internal review are not appropriate.

Copilot meeting preparation summarizes context from records and emails and can suggest questions to ask. A macro that counts emails, a lunch reminder and a blank agenda template do not provide that context.

Grounded answers use configured knowledge sources and show where information came from, which improves accuracy. Ungrounded guesses, personal notes alone and random colleague calls are unreliable.

A Copilot Studio agent can answer from knowledge sources and hand complex cases to sellers in Dynamics 365 Sales. Printed FAQs, yearly spreadsheets and voicemail do not provide AI answers with handoff.

Reviewing drafts ensures accuracy and tone, and keeping data in approved systems follows data policy. Sending unread messages, pasting confidential data into public tools and hiding required AI disclosures all create risk.

AI-based opportunity scoring uses historical and current signals to estimate likelihood of closing. Past closed lists, typed targets and printed charts do not predict outcomes for individual opportunities.

Sales forecasting combines AI predictions with commitments and allows managers to adjust. A copied forecast does not reflect current pipeline, optimism is not evidence, and personal notebooks are not shared forecasts.

Factor analysis of opportunity fields and activity signals shows which factors are associated with wins. Random selection, colour lists and licence counts do not explain win factors.

Filtering by days since last activity highlights stalled opportunities, which allows timely follow-up. Name sorting, holiday calendars and random samples do not identify stalled deals.

Alerts on significant probability changes prompt timely action. Monthly printed reports, static charts and yearly manual checks are too slow to react.

Current stages and close dates, and recorded activity, give the model accurate signals about progress. Blank close dates, duplicated opportunities and deleted closed deals all degrade the training and scoring data.

Next best action recommendations surface suggested steps inside the sales workspace based on the opportunity context. Printed lists, plain emails and holiday calendars do not provide contextual recommendations.

Predictive lead scoring ranks leads by likelihood to convert using historical data, which helps focus effort. Alphabetical sorting, random assignment and identical emails do not prioritize by likelihood.

Scoring details show the contributing factors, which helps representatives understand and trust the score. Colour preferences, page counts and office time zones do not explain the score.

Reviewing outcomes across groups and monitoring over time helps detect bias and maintain fairness. Ignoring patterns, removing transparency and uniform scores without review do not address bias.

Historical lead records and their conversion outcomes in the CRM provide the labelled examples a scoring model learns from. Random pages, vacation records and coffee sales are not relevant training data.

Choosing the attributes or signals the model uses lets the business align scoring with its policy and goals. Using all fields with no choice, random scores and name length do not reflect the requirement.

Comparing predicted scores with actual outcomes measures model accuracy over time, which shows whether it performs well. Lead counts, dashboard colours and email counts do not measure accuracy.

Models perform best on data similar to their training data, so poor results in a new region suggest retraining on representative data. Regional quality claims, design limits and blaming sellers without evidence are not the explanation.

Conversation intelligence analyses recordings and transcripts to surface talk ratios, mentions and other signals for coaching. Brochures, manual timers and room bookings do not analyse conversations.

Consent must be obtained and recorded according to policy and applicable law before recording and analysis. Recording without informing, public sharing and disabling consent checks breach that requirement.

Copilot needs a transcript and permission to access the meeting content, so transcription and access must be set up. Printed copies, phone lines without recording and static slides do not provide a transcript.

Topic and keyword trends from conversation intelligence summarize recurring themes across calls. Room bookings, weekday counts and passwords do not show customer topics.

Data subject requests must be handled through the defined process, including deleting recordings and derived data where required. Ignoring requests, sharing recordings and keeping them in personal folders breach privacy obligations.

Conversation intelligence reports talk-to-listen ratios from recordings, which supports coaching. A stopwatch, a file colour and a room name do not measure call dynamics.

A prioritized daily view uses insights to show what needs attention first. Holiday calendars, vendor lists and phone books do not prioritize work.

Competitors are recorded as linked entities so the team can analyse competitive losses and wins. The other options are unrelated to competition.

Sales processes guide sellers through stages with recommended steps, which standardizes the sales approach. Passwords, backups and vendor invoices are not sales guidance.