1. Experience and Expertise

Guideline: Buyers should assess the supplier’s experience and their balance of AI-relevant skills (e.g., data science, NLP) and research expertise.

Answer: While Flowres has been a leader in the research domain for years, our journey in bringing AI into the mix is rooted in our extensive experience and deep understanding of researchers’ needs. Founders of flowres are having more than 50 years of combined experience in qualitative research. flowres was designed to enhance productivity by addressing the pain points we saw firsthand in qualitative research, especially in handling vast, unstructured data.

As AI matured and became accessible, we integrated it into our platform thoughtfully, focusing on practical, high-impact areas like automated transcription, Q&A, and hypothesis testing. Our expertise in research underpins how our AI is deployed, making it highly relevant and directly beneficial for qualitative research applications. We’ve refined our AI solutions through real-world client feedback ensuring that the tool not only provides automation but also enhances research quality and insight speed

2. Positive Impact of AI

Guideline: Suppliers should explain how AI addresses research issues and improves the decision-making process.

Answer: AI comes with several possibilities and its almost impossible to list down all the areas where AI can impact research. Since we are focused on qualitative research, here is our list of areas that AI will impact going forward:

Automated Transcription in English and other languages: AI can transcribe interviews and focus group discussions quickly and accurately, saving time and reducing manual effort. While the AI transcriptions are far from perfect, we have seen a huge improvement in English to English transcriptions.

Enhancing Data Analysis for accuracy and thoroughness: AI algorithms are great at identify patterns, themes, and trends within large unstructured datasets, helping researchers uncover insights that might be missed manually. Besides, we believe AI will soon handle time-consuming tasks like coding and categorizing data, allowing researchers to focus on interpretation and strategy.

Real-Time Analysis: Unlike humans, AI can process data as it comes in, enabling fast toplines and quick analysis. This opens up avenues for researchers to adjust methodologies or explore new areas quickly.

Secondary Research during proposal and report writing: AI can sift through vast amounts of online content, reports, and articles, identifying relevant information and summarizing it efficiently. This allows researchers to gather comprehensive background data quickly, enhancing the quality and relevance of studies.

Document Drafting: Tools powered by AI can assist in drafting essential documents like discussion guides, screeners, and interview questions. By analyzing previous projects and industry language, AI ensures documents are aligned with best practices and specific project objectives, streamlining preparation and reducing time spent on these tasks.

Generating Concept Images: AI-generated images help researchers create visuals for concept testing, making it easier to present ideas in a tangible form without needing a dedicated designer.

Brainstorming and Ideation: AI-driven tools support brainstorming by suggesting new angles, identifying trends, or generating preliminary ideas. This accelerates the creative process and fosters innovative thinking, which is particularly useful in the early stages of research.

Using AI to Critique: Researchers can Anthropomorphize AI and ask AI to become a client, a visual critique, a colleague etc and collaborate with AI to improvise the quality of discussion guides, screeners, report presentations etc.

Flowres.io helps qualitative researchers accelerate the analysis of large, unstructured datasets, such as interview transcripts and survey responses.

Basically the applications for AI are vast in way we do research and possibly we haven’t discovered all of them. As AI tool developers, we are currently focusing on the largest impact areas like transcription and analysis at the moment without losing the sight of new possibilities that AI bring to the table.

3. Challenges in AI Use

Guideline: Suppliers should provide insights into the practical challenges faced and lessons learned in AI deployment.

Answer: Ensuring that AI-generated analysis remains aligned with the nuanced needs of researchers in terms of data audit trails, and validating output. A Use of AI for deductive coding, where we have a code list and want to assign them to the qualitative data (transcripts). AI is unable to take context into account and assigns many codes to many transcripts' cues. Staying on top of developments and releases from the big 3 in LLM’s (OpenAI, Gemini, Anthropic) to ensure optimal and relevant output with appropriate boundary conditions. We continue to test and update the system to improve performance in diverse scenarios.

4. Explainability

Guideline: The AI’s functionalities should be explained clearly to help stakeholders understand the process and benefits. Answer: flowres is a web app that helps researchers transform their generic online meeting platforms into a power-packed qualitative research hub. It reduces friction in the journey from data collection to data analysis to reports.

AI has been applied thoughtfully across the platform in the following ways –

Automated transcription – automatically transcribes online IDI’s and FGD’s with time stamps, speaker identification, and PII removal. This is available for English and 19 other languages.

Qualitative data analysis – reads through the transcripts, recognizes patterns or themes, and organizes the data to make it easier for researchers to analyze and interpret. Key functionalities include thematic grouping, frequency distribution, hypothesis generation, key driver analysis and more.

Citations – reviews and links the analysis to actual parts of conversations in transcripts and video’s

5. AI Model Details

Guideline: Buyers need to know if the AI is internally developed or involves third-party components for risk assessment.

Answer: flowres, is an application layer built on top of commercially available, enterprise grade LLM’s, mainly Gemini and OpenAI. We settled on the usage of these two models for the following reasons.

- Performance and Accuracy
These are extensively trained on massive, high-quality datasets, including professionally curated data across numerous domains. This typically results in higher performance and accuracy for complex, nuanced language tasks, especially important in market research where interpretation needs to capture subtlety and context.

- State-of-the-Art Updates and Innovation
Proprietary LLMs tend to be at the cutting edge of NLP research, often incorporating the latest advancements in efficiency, interpretability, and capability. For example, OpenAI and Gemini frequently release updates and improvements that reflect new research in language understanding and bias mitigation, which can benefit tools that rely on up-to-date AI capabilities.

- Robustness and Stability
Market research requires consistency and reliability. Proprietary models from established providers typically undergo extensive validation, stress-testing, and refinement to ensure that they handle a wide range of inputs robustly.

- Integrated Features and Ecosystem Support
Proprietary models often come with additional tools, APIs, and ecosystem support that enhance functionality. OpenAI, for instance, offers various API options that include built-in safety mechanisms and content filters, along with support for prompt engineering, making them easier to implement quickly and securely in platforms like Flowres.io.

- Data Privacy and Security
Proprietary models typically offer enterprise-grade privacy options, often including Data Protection agreements that prevent them from using sensitive data to train the models. This enables client organizations to meet stringent data security requirements without additional setup.

- Community and Compliance Support
Proprietary providers often maintain dedicated support teams and documentation for compliance with regulations (e.g., GDPR), which can be critical in market research. Open-source models can be more challenging to adapt and monitor for legal compliance since the responsibility falls entirely on the organization to maintain and validate the model’s alignment with legal standards.

6. Algorithm Functionality

Guideline: Suppliers should explain how the algorithm processes data to meet the business objectives.

Answer: flowres.io brings together specific functionality for qualitative data analysis at the application level, and leverages the LLM’s innate capabilities. The LLM and Flowres interact as follows –

- Data Processing and Pre-analysis
flowres organizes and pre-processes the data that users upload, such as interview transcripts and Excel notes for the LLM to analyze. This might include identifying the question and response, removing unnecessary elements, so the LLM can focus directly on the meaningful content.

- Visualizing and Summarizing Data
flowres organizes the LLM’s findings into formatted summaries and grid visualization, which makes it easy for researchers to digest and interpret the results. This visual organization lets researchers see trends and key points without needing to read through every transcript manually.

- Customizing and Fine-tuning Results:
flowres allows researchers to interact with the LLM’s outputs, making adjustments to themes or categories if needed. Researchers can refine the analysis to better align with the study’s specific goals or explore findings in more depth by asking follow-up questions or re-running certain analyses with a new focus. They can also organize analysis visually for individual elements as per their standards and requirements.

The underlying LLM’s detect common themes, patterns, and anomalies by analyzing word frequency, context, and sentiment. Client data is handled with strict data governance protocols, and client’s data is never used for AI training.

7. Validation and Bias

Guideline: Suppliers should detail their validation processes, biases management, and measures for ensuring fit-for-purpose results.

Answer: As a process, the team behind Flowres validates output accuracy by frequently checking AI-generated insights against expert human analysis. We have in-house researchers who regularly run these checks.

The platform itself restricts the data available for analysis to the transcripts uploaded, and has functions that ensure this is strictly adhered to. In the event of a prompt that potentially requires information from other sources, the system is programmed to ask the researcher to reframe the question.

Our team regularly reviews system outputs, as we believe human oversight is key to ensuring the AI remains reliable and fit for purpose. And all question prompts are required to be user defined, with templates being provided for more effective prompting.

8. Limitations of AI

Guideline: Transparency around limitations and mitigation strategies should be provided to manage expectations.

Answer:

- Potential for Inaccurate or Biased Outputs
These models can generate responses that are plausible but incorrect or reflect inherent biases present in their training data. In market research, relying solely on AI-generated insights without human validation can lead to misguided conclusions. At Flowres, we ensure that our users are aware of these potential problems, and we do provide the capability to switch the underlying model if any bias is reported by the user.

- Lack of Domain-Specific Expertise
While GPT-4 and Gemini are trained on diverse datasets, they may lack the specialized knowledge required for niche market research sectors. This limitation can result in superficial analyses that overlook critical industry-specific nuances. To mitigate this, we have provided capability for users to upload project specific information such as domain terminology, discussion guides and briefing notes to improve the context that the LLM operates in.

- Limited Access to Proprietary or Real-Time Data
These models do not have access to proprietary databases or real-time information unless explicitly provided during the interaction. This constraint can affect the timeliness and relevance of the insights generated, which is vital in dynamic market environments.

- Interpretability and Explainability Challenges
The complex nature of these AI models can make it difficult to understand how they arrive at specific conclusions. This lack of transparency can be problematic in market research, where clear rationale behind insights is necessary for informed decision-making. To mitigate this, we provide a reverse look-up to the actual parts of the conversation that have been referenced when writing observations, inferring themes or drafting summaries.

9. Ethics and Duty of Care

Guideline: The supplier should explain how they address potential negative consequences of AI use on people.

Answer: We continuously evaluate the ethical implications of our technology, especially when dealing with sensitive research topics like health, or maintaining consumer privacy.

We are ISO 27001 certified, MRS company partner and ESOMAR members. We follow the Guidelines mentioned in the MRS Code of Conduct, and in the ICC/ESOMAR International Code on Market, Opinion and Social Research and Data Analytics. As part of these guidelines, we ensure that

- no questions apart from those related to the conversation in the transcripts is analyzed by the LLM - all speaker identification is stripped of PII, so R1, R2 etc. replace respondent names - audio only live streaming for sensitive conversations, the stream itself is user consent driven - unless requested by clients, all interview data is purged from the servers after 3 months - access to interview data is tightly controlled

10. Transparency in AI Use

Guideline: There should be transparency in communicating when AI is driving outputs or decisions.

Answer: As a platform, flowres leverages AI only where necessary, it is clearly labelled as such for AI usage in data analysis. Users are informed, and they have the ability to monitor the data journey throughout the platform.

AI is used through out the platform for these purposes –

Automated transcription – Clients and users are made aware that automated transcripts are leveraging commercial API’s, and encourage use of human proofreading to improve accuracy.

Qualitative Data Analysis – This is a GenAI app, and users are aware that the outputs from this application depend on the prompts provided by the users.

Citations – These are provided as a single window validation mechanism for the QDA output.

11. Defined Ethical Principles

Guideline: Ethical principles should govern AI behavior, with human oversight to ensure alignment with these principles.

Answer: flowres follows a robust set of ethical principles to guide AI behavior, emphasizing data privacy, non-discrimination, and accountability. Our platform is built on established, open-source LLMs with defined ethical standards, and we reinforce these through human validation of all AI outputs. Flowres.io also offers a researcher-led validation service to ensure ethical compliance, ensuring that AI outputs align with human-defined ethical frameworks, especially in sensitive or impactful research contexts.

Ethical Principles as Our Foundation
At Flowres.io, our ethical principles guide every aspect of our platform’s design and operation. We prioritize Transparency, Privacy, Fairness, Safety, and Accountability. These principles shape how we integrate and apply AI, ensuring that each AI-driven insight aligns with responsible research practices.

Ensuring Human Oversight
We emphasize human oversight as an essential component of ethical AI use. Researchers using Flowres.io are encouraged to validate and interpret AI outputs, combining human expertise with AI capabilities to achieve meaningful results. This approach prevents over-reliance on AI alone, making sure that ethical principles are upheld throughout the research process.

Continual Ethical Evaluation
Flowres.io actively gathers user feedback and reviews platform practices to uphold our ethical standards. By combining this feedback with relevant guidelines published time to time by ESOMAR and MRS, we keep our platform aligned with evolving ethical expectations, ensuring that our AI is used in a way that respects human values and promotes responsible research.

12. Human Oversight

Guideline: Suppliers should describe how human involvement is used throughout the AI process, including feedback loops and validation.

Answer:

Human-Informed AI
flowres.io places human expertise at the center of our AI-driven insights. While the platform automates repetitive tasks, researchers retain full control to validate, interpret, and refine AI outputs, ensuring that every insight aligns with the study’s objectives and ethical standards.

Feedback and Refinement Tools
Our platform includes tools that allow researchers to review and adjust AI-generated results. This oversight ensures that insights are nuanced, relevant, and compliant with ethical research practices. flowres.io also provides resources, such as best-practice guides and webinars, to help users combine AI insights with their expertise.

Ethics in Practice
With flowres.io, human oversight is not just encouraged; it’s built into the platform’s design. We believe that responsible innovation in AI means empowering researchers to blend AI outputs with their knowledge, creating insights that are both ethical and impactful.

13. Data Quality

Guideline: Suppliers must ensure high-quality, representative data for reliable AI outcomes.

Answer:

Ensuring Data Relevance
flowres.io uses commercial LLMs like OpenAI and Gemini, which are trained on vast, diverse datasets. We carefully select these models for their ability to deliver accurate, relevant results, supported by the rigorous data processing standards of our LLM providers.

Focused Data Use
To ensure quality outcomes, flowres.io only processes user-uploaded data in context-specific applications. We prioritize accuracy and relevance by enabling researchers to refine and contextualize AI outputs, making adjustments that align insights with specific research objectives.

Continuous Improvement
We maintain high standards for data quality by actively gathering user feedback, which informs future platform updates and enhancements. This approach helps Flowres.io stay responsive to research needs, providing outputs that meet the expectations of our diverse user base.

14. Data Lineage

Guideline: Suppliers should clearly track and communicate the sources and processing of the data used in AI.

Answers:

Clear data tracking
All qualitative data comes from clients, and is tightly controlled once it enters the system. Clients are able to directly trace output to underlying data, which is clearly marked and identified at the time of input. Further, timestamps are maintained in the databases to record times of data upload, export times etc.

User Control and Transparency
We provide researchers with visibility into data processing, enabling them to understand and verify the data journey within Flowres.io. By keeping a record of all data inputs and processing steps, we offer researchers the assurance that data lineage is maintained at every stage.

Outputs at each stage of data processing
We currently also provide an auxiliary Content Analysis service using our proprietary CAQDAS. We provide raw coded output, which provides traceability coupled with outputs that make raw data traceable. This can be optionally used by our clients for analysis using our Gen AI analysis application, and this has complete traceability.

15. Privacy Notice

Guideline: Suppliers should ensure compliance with data protection laws through clear privacy notices.

Answer:

Our Privacy Commitment
flowres.io is dedicated to transparent data handling and privacy protection. Our privacy policy, details our approach to data collection, processing, and storage. We comply with GDPR and ISO 27001 standards to ensure user data is secure and handled responsibly.

Privacy policy Accessibility
We encourage all users to review our privacy policy to understand how flowres.io protects their data. This notice outlines our data management practices and underscores our commitment to upholding the highest standards of data privacy in every interaction with the platform. flowres.io’s privacy policy is publicly available and outlines our approach to handling personal data, including data collection, processing, and storage. We comply with all relevant data protection laws, including GDPR, and ensure that user data is securely managed.

16. Data Protection Measures

Guideline: Suppliers should outline compliance steps and risk mitigation, including obtaining consent when necessary.

Answer:

GDPR and ISO Compliance
flowres.io adheres to GDPR and ISO 27001 standards, implementing strict data protection measures to safeguard user data. We have robust data access controls, encryption protocols, and auditing practices in place, ensuring data privacy throughout its lifecycle on our platform.

Risk Mitigation and Consent
We evaluate potential privacy risks as required by data protection regulations and obtain consent where necessary, ensuring our compliance aligns with industry standards. Our Data Processor agreements with OpenAI and Gemini further protect user data, keeping it confidential and restricted to research purposes only.

17. Resilience to Attacks

Guideline: Suppliers should discuss the security frameworks they use to ensure the system’s resilience and integrity.

Answer:

Security Standards and Protocols
flowres.io follows ISO 27001 standards, incorporating best practices in cybersecurity to protect against adversarial attacks and system disruptions. We utilize advanced firewall protections, encryption, and continuous monitoring to ensure system resilience and data integrity.

Regular Assessments and Updates
Our platform undergoes routine security assessments to identify and mitigate potential vulnerabilities. By proactively addressing security risks, Flowres.io remains prepared to handle disruptions and protect user data, providing a stable and reliable platform experience.

Commercial LLM preparedness
We only leverage commercially available, tested LLM’s. OpenAI and Gemini employ several strategies to bolster the resilience of their AI systems against adversarial attacks, noise, and other potential disruptions. These strategies include:

  • Adversarial Training
  • Robustness Testing
  • Security by Design
  • Adherence to Information Security Frameworks and Standards
  • Continuous Monitoring and Response
  • Collaboration and Transparency

18. Data Ownership

Guideline: Suppliers must clearly define and communicate ownership and usage rights for the input data.

Answer: flowres.io clearly defines data ownership from the outset, ensuring that clients retain full ownership and intellectual property rights over their data. Our platform processes data solely for the agreed-upon purposes, with usage permissions explicitly defined and documented.

We communicate data ownership terms clearly, keeping clients informed of how their data is handled, used, and protected. Flowres.io is committed to respecting user data rights, giving clients confidence and control over their information.

19. Data Sovereignty

Guideline: Suppliers need to clarify any limitations imposed on the use of the data by its owners, including regional or legal restrictions.

Answer:

Strict Data Use Limitations
flowres.io ensures that user data is only used for its intended research purposes. Our Data Processor agreements with Gemini and OpenAI explicitly restrict any secondary use of data, maintaining data confidentiality and integrity.

Compliance with Regulations
flowres.io aligns with data handling regulations such as GDPR, and data privacy requirements in acts such as HIPAA. Users retain control over their data’s usage, with flowres.io’s platform adhering to strict limitations on processing and access.

20. Ownership of Outputs

Guideline: Ownership of the AI-generated outputs, including intellectual property rights, should be clearly defined.

Answer:

Client Ownership of AI Outputs
Flowres.io is transparent about ownership rights, ensuring that clients retain full ownership of all AI-generated outputs. We clearly define these rights in our terms of service, allowing clients to use, share, and distribute insights produced by the platform.

Intellectual Property Clarity
Our commitment to clarity in ownership allows researchers to confidently use AI-generated outputs as they see fit. Flowres.io’s policies ensure that all insights remain the property of the client, giving them full freedom over the resulting data and analyses.