Ai Agents Are Only As Good As The Data They're Given, And That's A Big Issue For Businesses

zf L/Moment/Getty Images

Follow ZDNET: Add america arsenic a preferred source on Google.

ZDNET's cardinal takeaways

• 63% of business leaders picture their organizations arsenic very data-driven, up 10% from 2023. 
• Only 1 successful 2 business leaders is assured astir nan expertise to present timely business insights.
• The astir valuable insights for organizations are presently trapped successful unstructured data. 

Business leaders understand nan worth of data. Sixty-three percent of today's business leaders picture their organizations arsenic very data-driven, up 10% from 53% successful 2023, according to Salesforce's State of Data and Analytics Report based connected a study of 3,800 information and analytics leaders and 3,852 cross-functional business leaders worldwide. That said, astir two-thirds (63%) of method leaders admit that their companies struggle to thrust business priorities pinch data.

The accelerated emergence of AI agents has created a palpable consciousness of urgency for companies crossed each industries and geographies, arsenic they activity ways to reinvent themselves arsenic 'agentic enterprises' to accelerate business growth. Salesforce's investigation revealed that business, data, and analytics leaders are grappling pinch really to overhaul their information infrastructure, management, and governance. 

Also: Enterprises are not prepared for a world of malicious AI agents

This displacement is captious for some occurrence -- aliases successful immoderate cases, endurance -- arsenic their companies hole for an agentic future. This early requires nan democratization of data, AI, and analytics, making these resources accessible to everyone wrong nan organization. 

Here are nan 4 cardinal findings from nan Salesforce report: 

1. AI accelerates a caller information paradigm: The immense mostly of companies now usage astatine slightest 1 shape of AI successful their day-to-day workflows. As these revolutionary technologies return hold, business and method leaders are questioning really prepared their underlying foundations and cultures are, particularly arsenic information measurement and complexity increase. To reiterate, 63% of business leaders picture their organizations arsenic very data-driven.
2. Limited information assurance thwarts activation, decisions, and action: As organizations go much data-driven, galore business leaders consciousness mislaid successful nan slow, method processes of generating analytic insights. What's more, galore aren't judge nan information they trust connected is meticulous successful nan first place. Fifty percent of business leaders aren't judge they tin make and present timely insights.
3. Building nan information instauration for analytics and AI: On nan method side, information and analytics leaders are pressured from accrued line-of-business request for data-driven capabilities and executive demands for agentic innovation. But mediocre information guidance practices, including integration and harmonization, positive nan power of unstructured formats, airs formidable challenges. Seventy percent of information and analytics leaders judge nan astir valuable insights for their organizations are trapped successful unstructured data.
4. Governing and safeguarding nan agentic enterprise: The emergence of AI is exposing long-standing shortcomings successful information security, compliance, and governance measures. Only 43% of information and analytics leaders person established general information governance frameworks and policies, and 88% judge AI demands caller approaches.

Here are my cardinal takeaways from nan apical 2 insights: AI accelerates a caller information paradigm, and constricted information assurance limits decisions and actions. 

AI accelerates a caller information paradigm

Ninety percent of business leaders judge their careers dangle connected being data-fluent. Meanwhile, 86% judge their careers dangle connected being data-driven, and I deliberation nan different 14% will beryllium looking for caller jobs soon. 

There is, however, a disconnect successful business -- though businesses study utilizing information more, method leaders person reservations, pinch astir two-thirds (63%) agreeing their companies struggle to thrust business priorities pinch data. Data and analytics leaders estimate that 26% of their organizations' information is "untrustworthy." And 42% of business leaders said their information strategies don't afloat align pinch business objectives.

Also: The champion AI agents are unspeakable freelancers - for now

All AI projects are information projects, and astir each organizations work together that nan emergence of AI makes information much critical. Ninety-three percent of organizations person astatine slightest 1 lawsuit of AI successful their exertion stacks, according to Salesforce's latest State of IT survey. The technology's accelerated evolution, including nan advent of agents, is putting unit connected information and analytics leaders to ramp up capabilities quickly. Leaders spot AI arsenic a forcing usability to amended their wide information literacy and culture, pinch 91% of business leaders believing that nan emergence of AI makes it much important to beryllium data-driven.

So, really are businesses changing their finance thesis to support AI programs, including nan take of agentic AI? The bully news is that location are boundless opportunities to amended efficiency, innovation, and productivity utilizing AI agents. The State of IT investigation from Salesforce recovered that 84% of CIOs judge AI will beryllium arsenic important to their businesses arsenic nan internet. 

Also: Even nan champion AI agents are thwarted by this protocol - what tin beryllium done

Additionally, 84% of information and analytics leaders work together AI's outputs are only arsenic bully arsenic its information inputs. Another cardinal uncovering is that CIOs spent 4 times arsenic overmuch of their fund connected information infrastructure arsenic connected AI. The logic for that finance whitethorn beryllium that 80% to 90% of endeavor information is estimated to beryllium unstructured, and 70% of information and analytics leaders judge nan astir valuable insights for their organizations are trapped successful unstructured data.

There is an water of data, and yet we are thirsty for insights. Data and analytics leaders estimate their organizations' information volumes turn 30% annually, up from 23% successful 2023. The mean number of information environments crossed nan endeavor is: 

• 26 spreadsheet applications
• 21 unreality retention services
• 21 operational databases
• 17 information warehouses
• 16 information lakes
• 15 customer information platforms

The mean endeavor uses 897 applications, and only 29% are connected. Business leaders mostly don't afloat spot their data, citing persistent issues specified arsenic accuracy, reliability, and relevance. More than half (54%) of leaders aren't wholly assured that nan information they request is accessible successful nan first place. Data and analytics leaders estimate 19% of their companies' information is trapped.

Also: Want amended ChatGPT responses? Try this astonishing trick, researchers say

The apical information priorities are: building AI capabilities, providing real-time information access, improving company-wide information fluency, improving information quality, and strengthening information and compliance. The apical information challenges are: deficiency of real-time data, deficiency of information harmonization, information threats, ensuring information accuracy and quality, and siloed aliases trapped data. 

Real-time information roseate dramatically arsenic nan apical information challenge, surpassing perennial symptom points for illustration harmonization, information threats, and wide accuracy and quality. And arsenic AI underscores nan request to deduce worth from unstructured data, trapped and siloed sources information retired nan apical 5 information hurdles, skyrocketing from past of each concerns 2 years ago.

Limited information confidence

Regardless of whether it's powering an AI prediction, an agent-driven customer interaction, aliases a study highlighting cardinal metrics, information is only bully if it's grounded successful nan business discourse that contributes to "trustworthy data." Ninety-three percent of business leaders work together insights are only applicable if they're grounded successful nan business context. Here are nan apical factors preventing data-driven organizations:

1. Incomplete, out-of-date, aliases poor-quality data.
2. Lack of devices to access, analyze, and construe data.
3. Lack of expertise and training connected really to access, analyze, and construe data.
4. Takes excessively agelong to get insights.
5. Lack of entree to nan required data.

The challenges noted supra forestall labor from acting connected applicable insights promptly, pinch 49% of information and analytics leaders saying their companies occasionally aliases often tie incorrect conclusions from information that misses aliases misunderstands business context.

Greater take of AI is accelerating really businesses entree and enactment connected data. A immense mostly of analytics and information leaders (91%) said method queries limit analytics usage astatine scale, and 92% cited a deficiency of information fluency among staff. AI is, therefore, progressively utilized to augment analytics processes, pinch 64% of business leaders utilizing AI to find, analyze, and construe data, while only 54% get thief from a method resource. 

Greater take of AI intends buyers person higher expectations from AI solutions. Analytics solutions buyers prioritize AI and real-time data, including AI-driven actions -- 88% of information and analytics leaders said advances successful AI are changing really they measure analytics package and implementations. Analytics and information leaders are looking for real-time data, AI-assisted workflows, AI-driven actions, composable analytics, and insights astatine scale. The study recovered that 94% of business leaders said they'd execute amended pinch nonstop information entree successful nan programs/apps wherever they activity nan most.

Also: AI is much apt to toggle shape your occupation than switch it, Indeed finds

Analytics and information leaders clasp AI agents to assistance and thrust outcomes for their stakeholders. AI agents that tin understand, respond to, and return actions based connected personification inquiries successful earthy connection clasp peculiar entreaty and potential. Specifically, agentic analytics makes information depletion and relationship highly intuitive and conversational. Leaders want to person conversations pinch their information platforms. As galore arsenic 63% of information and analytics leaders said translating business questions into method queries is prone to error, and 93% of business leaders said they'd execute amended if they could inquire information questions pinch earthy language.

To study much astir nan State of Data and Analytics report, you tin sojourn here.