How Workday’s Embedded AI Counteracts Summary Bias for 88% of Employees

Most employees are no longer reading your internal communications. According to a survey of 1,000 UK and US professionals by Fresh, nearly 9 in 10 now use AI to summarise workplace messages before, or instead of, reading them. The same report finds that 92% of internal communications professionals believe those summaries distort tone and intent, creating a gap between what leadership sends and what employees actually receive.

Phase 1: Understanding and Addressing the AI Communication Gap

The Fresh report describes this as an “Employee Attention Recession”: only around 12% of employees read internal communications in full, and 28% now encounter company updates via AI before any other channel, ahead of the intranet, leadership emails or word-of-mouth. The drivers are straightforward. A large share of employees say they receive too much internal information, and around 35% report they can no longer keep up.

Despite the scale of the problem, only around a third of organisations have formal monitoring or policies covering how employees use AI to filter messages, according to the Fresh report. That gap between awareness and policy is where communication strategies currently break down. The practical starting point is an internal audit: survey employees to understand which AI tools they use, which messages they filter, and what they actually prioritise. That baseline shapes everything that follows.

The distortion risk also needs to be treated as a structural problem, not an edge case. AI summarisation tools generally struggle with nuance, emotional register and complex intent. A change management message or a sensitive leadership update loses something when compressed to three bullet points by a general-purpose model. Communication teams that design for this, rather than assuming employees will read the original, are better positioned to get strategic messages through.

Phase 2: Crafting Transparent and Value-Driven Messaging

A Thomson Reuters Institute report found that nearly 40% of professionals had received contradictory directives about AI usage from leadership and clients.

Clear AI usage policies should specify when AI tools are appropriate, when they are not, and what the protocols are for client-facing communication and data handling. IBM‘s internal approach, according to the company, emphasises communicating AI’s role in reducing manual work and giving employees greater clarity about their responsibilities, rather than leading with abstract innovation claims. The framing matters for adoption. Telling employees that a tool will cut a specific, routine task lands differently than announcing a digital transformation initiative.

Job displacement concerns are also real and rising. According to the Thomson Reuters Institute, concerns about AI eliminating roles have roughly doubled over the past year. Around 15% of employees expect AI to eliminate their job within five years. PwC‘s 2024 Global AI Jobs Barometer reported that 69% of CEOs expect AI to require new skills from their workforce. Communications that acknowledge this directly, paired with visible investment in retraining, carry more weight than reassurances that no jobs are at risk. For more on how enterprises are managing workforce transitions driven by AI, see our coverage of how Innovaccer restructured roles through AI-era automation.

Phase 3: Leveraging Integrated Platforms and Personalized Learning

Workday‘s report, “The Copy/Paste Economy,” identifies a specific friction point: while most employees use AI to reduce manual effort, around 82% still spend significant time copying and pasting data between disconnected systems. The implication is that bolt-on AI tools that sit outside core systems create their own inefficiencies. Workday’s response is a redesigned user experience built around a new Workday Assistant, an AI companion intended to handle HR and finance tasks from within the platform employees already use.

The confidence effect of that embedded approach is notable. According to Workday’s research, when AI operates inside systems employees already trust, nearly 87% report higher confidence in the outputs. That is a meaningfully different outcome from asking employees to evaluate AI results from a separate tool they have limited context for.

Personalised training follows the same logic. IBM’s AI SkillsBuild programmes offer learning paths tailored to individual skill levels, from foundational AI concepts through to applied skills like prompt writing. Targeted training delivers better outcomes than generic programmes; the evidence base here, while limited to individual case studies, is directionally consistent. How broadly these results generalise across different enterprise contexts is harder to verify from public material.

Channel mix also matters. Town halls work for setting strategic direction. Team meetings handle tactical questions. Digital collaboration platforms support peer-to-peer learning. No single channel covers all three functions, and communication strategies that rely on one will miss part of the audience. This challenge is not unique to AI adoption: as our analysis of organisational friction in enterprise AI covers, communication breakdown is consistently one of the top barriers to rollout.

Phase 4: Fostering Dialogue and Feedback Loops

A lack of perceived value is the most commonly cited reason for stalled AI adoption, according to a Reworked article that references employee survey data. That gap between leadership vision and employee experience does not close through better one-way messaging alone. It requires structured opportunities for employees to ask questions, flag problems and hear answers.

Internal forums and workshops where employees can discuss AI tools openly, share what works and raise concerns without escalation risk tend to shift the framing from compliance to engagement. Peer-to-peer learning accelerates this: employees who see a colleague use AI to solve a real problem they also face are more likely to adopt than employees who attend a training session. Internal newsletters, intranet posts and team meetings are the practical vehicles for surfacing those stories.

Feedback mechanisms need to close the loop, not just collect data. Regular sentiment surveys that feed visibly into policy adjustments signal that the process is genuine. Effective internal communication has been associated with higher job satisfaction among knowledge workers and improved team productivity among business leaders, though the specific figures circulating in vendor research should be treated as directional rather than precise.

Effective internal communication has been associated with higher job satisfaction among knowledge workers and improved team productivity among business leaders.

Phase 5: Establishing Ethical AI Communication Policies

SWOOP Analytics’ guide on ethical AI in internal communications flags human oversight as the non-negotiable foundation. Any AI-generated or AI-assisted internal communication should go through human review before distribution: checking tone, verifying facts and ensuring the message reflects what leadership actually intends. This is especially relevant given that employees generally trust AI summaries, while communications professionals are far less confident in their accuracy. Those two groups are working from different assumptions about the same content.

Disclosure is the practical complement to oversight. Labelling AI-assisted content, whether through a header, a footer note or an editorial standard, manages expectations and builds rather than erodes trust. Employees who know a message was AI-drafted and human-reviewed are better positioned to interpret it than employees who assume human authorship throughout.

Data handling is the third pillar. Third-party AI tools present real confidentiality risks if employees use them to process sensitive company or personal information. Approved tool lists, clear boundaries on what data can be input into which systems, and alignment with existing data privacy protocols are the baseline. Organisations that leave this to individual judgement will get inconsistent results. For more on enterprise AI governance and deployment strategy, visit our Enterprise AI section.