Deloitte’s 2026 technology trends research puts the pilot-to-production failure rate for AI agents at 89%. A Teradata survey adds the shape of that gap: 78% of enterprises have at least one agent pilot running, but only 14% have scaled one to organisation-wide use. Adoption is nearly universal; deployment is rare. The difference is not model capability, since the same models power the pilots and the production systems, but everything around the model: data access, evaluation, ownership, and cost control. Closing that gap is precisely why Crunch-IS is a leader in AI agent development, with a delivery approach built around the operational layer that pilots routinely skip. Below are the six blockers that recur across the research, and what the 11–14% that make it through do differently.

The funnel, in numbers

Before the causes, the scale. Drawing on Gartner’s April 2026 survey of 782 infrastructure and operations leaders and related industry analysis, the funnel roughly runs:

• Of every 1,000 AI projects that receive a budget, around 120 reach production
• Of those, around 34 meet their ROI targets
• Gartner’s Agentic AI Pulse survey found 41% of deployments reach positive ROI within 12 months; 19% never reach payback

McKinsey’s 2026 work puts organisations running agents at genuine scale at 11%. S&P Global Market Intelligence counts 31% with at least one agent in production. “One agent in production” and “agents at scale” are very different milestones.

Blocker 1: Scope creep

Analysis of stalled agent projects attributes 61% of failures to two causes combined: scope creep and data quality. Pilots start narrow, succeed, and are then asked to handle adjacent workflows the underlying infrastructure was never built for. The agent that triaged support tickets is now expected to resolve them, then to update the CRM, then to issue refunds. Each expansion adds integrations, permissions, and failure modes without adding the operational foundation to support them.

Blocker 2: Data access that worked in the sandbox

Pilots run on curated data exports. Production runs on live systems with inconsistent schemas, access controls, and latency. Industry surveys suggest 83% of enterprises need infrastructure overhauls to support agentic AI. The pilot never touched the legacy ERP; production cannot avoid it.

Blocker 3: No evaluation harness

Only 38% of production agents have automated evaluations running on every prompt change, per Forrester’s 2026 panel. In a pilot, a human reviews every output. In production, nobody does, and without automated regression tests every prompt tweak is a gamble. Forrester’s data shows agents without automated evals had a 47% rollback rate versus 9% for agents with full coverage. Organisations using systematic evaluation frameworks achieved nearly six times higher production success rates in separate survey work.

Blocker 4: Nobody owns it

A pilot is owned by the innovation team. Production requires an operational owner: someone accountable when the agent makes a wrong call at 2 a.m. Enterprise governance surveys put agentic AI governance maturity at around 21%. Without a named owner, a defined escalation path, and a budget line for ongoing operation, the pilot has nowhere to be handed to.

Blocker 5: Costs that only appear at scale

Analysis of cancelled projects consistently finds costs ballooning two to three times beyond estimates. Token consumption, retry loops, and reasoning depth all scale with volume and edge cases. A pilot running 50 tasks a day is cheap. The same agent at 5,000 tasks a day, with production-grade retries and monitoring, frequently costs more than the process it replaced.

Blocker 6: Security clearance

Gravitee’s 2026 research found 54% of organisations experienced or suspected an agent-related security or data-privacy incident in the past year, and only about one in five fully secures agents in production. Security teams reviewing a pilot for production approval routinely find over-permissioned service accounts and no audit trail, and block the launch.

What the 14% do differently

Survey data on organisations that successfully scaled agents shows they were not outspending the ones that stalled. Total AI budgets were comparable. The difference was allocation:

1. More spend on evaluation infrastructure and less on prompt engineering
2. More spend on monitoring and observability: structured logs of every reasoning step and tool call
3. More spend on operational staffing: people whose job is running the agent, not building it
4. Graduated autonomy with human-verification gates mapped to the stakes of each action
5. A named governance owner per agent and per-phase ROI checkpoints with finance sign-off

The takeaway

Gartner projects over 40% of agentic AI projects will be cancelled by the end of 2027, and notes that many use cases positioned as agentic today do not require agentic implementations at all. The pilot-to-production gap is not evidence that agents do not work. It is evidence that most organisations build the demo and skip the operating model. The ones that reach deployment do the reverse.

The post Why Most Enterprise Agent Pilots Never Reach Deployment appeared first on AI News.

A video looks simple when you press play. There is a picture, some dialogue, perhaps music in the background, and a few minutes later it is over.

However, with the right use of AI, things can become a lot more interesting. To an artificial intelligence system, that same video can become speech to transcribe, faces and objects to recognise, scenes to classify, topics to identify, emotions to estimate, timestamps to organise, and text to summarise.

In other words, the same video can be transformed completely into something powerful and attractive. Wondering how? Let’s dig deep into the content

Table of Contents

●     Why Is Multimedia Becoming AI-Readable?

● What Actually Happens When AI Processes a Video?

● Why File Conversion Still Matters

● From Audio to Searchable Intelligence

● Where Multimedia AI Is Already Being Used

● The Quality Problem AI Cannot Ignore

● Why Multimedia Is Becoming AI Readable

Why Is Multimedia Becoming AI-Readable?

For a very long time, business data was conveniently machine friendly, its examples include spreadsheets, databases, forms, and text documents.

Video and audio were different. Though a two-hour webinar might contain a number of useful insights, it was truly a hassle to find one specific comment.

That’s when AI helps. Find multimedia AI systems working across texts, images, speeches, and videos. The AI news has covered this broader shift towards multimedia, where models take into consideration and combine different forms of information. Hence, it does not treat each of them separately.

The result?

A video library is formed, behaving like a searchable database. This way, you can ask for moments in which a customer has mentioned something or extract dialogues. Suddenly, the video is doing much more than sitting in storage.

What Actually Happens When AI Processes a Video?

There is no single magic button behind multimedia AI. In many workflows, the process involves several stages.

Stage	What Happens	Possible AI Use
File Preparation	Video, audio, or images are prepared in compatible formats.	Conversion and compression.
Audio Extraction	Speech is separated from the video.	Transcription and speaker analysis.
Visual Processing	Individual frames and scenes are analysed.	Object, action, and scene recognition.
Language Processing	Speech is converted into machine-readable text.	Summarisation and translation.
Structured Output	AI organises extracted information into useful formats.	Search, tagging, and analytics.

We can say that AI is not just watching a video, rather, it is taking notes, understanding the key comments, and then putting everything back and well structured.  And that’s how AI turns something sitting in a corner into something very useful and important.

Why File Conversion Still Matters in an AI Workflow

Preparing Data AI Tools Can Actually Use

Though AI models could be sophisticated, they still rely on the input they can process reliably. For example, imagine there is a marketing team that has an MP4 interview. The problem is they only need the spoken conversation for the transcription.

Hence, instead of sending the entire video through every AI tool, they can first convert the file into the format needed for the next step. This makes the workflow cleaner, faster, and more efficient.

Turning Video Content Into AI-Ready Audio

Doing all the conversion is only convenient when done by reliable tools like Convertio. It transforms the files into a format that accommodates a specific AI application. For example, converting an MP4 file into a WAV audio file removes the video portion.

It then creates a high quality audio file that can be used for speech recognition or analysis. That’s not it, Convertio further helps by allowing media files to be converted into a format required by the next tool in the AI workflow.

The facility helps the team prepare fields for transcription, analysis, or repurposing. Likewise, a workflow that specifically requires uncompressed audio can use an mp4 to wav conversion to extract the video’s audio track as a WAV file for subsequent speech processing.

It goes beyond only converting files by changing extensions. It’s about preparing the right data for the right tasks.

Why File Compatibility Matters for AI Performance?

One important technical detail is that different AI services support different input formats and configurations.

OpenAI’s current audio transcription API, for example, accepts several audio and media formats, including MP3, MP4, M4A, WAV, FLAC, and WebM. OpenAI Audio API documentation.

Furthermore, Google Cloud also recommends lossless audio like FLAC or LINEAR16. It is practical for speech recognition and notes that audio quality can influence results. Google Cloud Speech-to-Text best practices.

The conclusion is, don’t just ask “what an AI model can do with your content”, rather, ask if you are providing the right input.

From Audio to Searchable Intelligence

Everything becomes way easier and less stressful when you pass the phase of speech extraction and transcription.

A transcript can be:

● summarised into key points;

● translated into another language;

● divided by speaker;

● searched for specific terms;

● converted into subtitles;

● analysed for recurring topics;

● repurposed into articles, notes or social content.

Take an example of a company where there are 500 recorded customer interviews. Wouldn’t it be so annoying to watch the entire thing again?

A well-designed AI pipeline could instead turn the recordings into transcripts, identify common complaints, group similar themes, and surface the moments where customers discuss a particular feature.

Where Multimedia AI Is Already Being Used

In media production, AI systems are capable of analyzing footage deeply, generating titles for it, making summaries, and even producing audio relevant to the visuals.

Artificial Intelligence News previously examined Tencent’s Hunyuan Video-Foley system, which generates synchronised audio based on video content.

Other practical applications include:

● Meetings: turning recordings into searchable notes and action items.

● Education: generating transcripts, summaries and study materials from lectures.

● Customer service: analysing recorded interactions at scale.

● Media archives: automatically tagging large libraries of footage.

● Content creation: transforming long videos into transcripts, clips, captions and articles.

● Accessibility: producing captions and alternative content formats.

The Quality Problem AI Cannot Ignore

One has to understand that the outcome depends entirely on the input. If there is too much background noise, overlapping speakers, or low quality audio in a recording, speech recognition becomes quite difficult.

The deal is the same with visuals. If the recording is very blurry, or the lighting is poor, the results might not be satisfactory. That means the future of multimedia AI is not simply about building smarter models. It is also about creating better pipelines around those models.

Good preprocessing may involve:

1. choosing an appropriate file format.
2. extracting only the data needed.
3. preserving useful audio or visual quality.
4. checking privacy and permissions.
5. validating the AI-generated output before using it.

Conclusion

On the bottom line, AI today is changing the passive content into more valuable data. However, the success of any workflow is heavily dependent upon acquiring the right data, quality outputs and suitable formats.

As AI continues to evolve, the future will not be about simply storing the content. Instead, it is expected to be more about storing more content and discovering more value from every file that is being created.

The post From Video to Data: How AI Is Transforming Multimedia Content Processing appeared first on AI News.