DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
This Office Action is in response to correspondence filed 30 July 2026 in reference to application 18/940,609. Claims 1-20 are pending and have been examined.
Response to Amendment
The amendment filed 30 July 2026 has been accepted and considered in this office action. Claims 1, 9, and 16 have been amended.
Response to Arguments
Applicant’s arguments with respect to claim(s) 1-20 have been considered but are moot because the new ground of rejection does not rely on any reference applied in the prior rejection of record for any teaching or matter specifically challenged in the argument.
Claim Rejections - 35 USC § 103
The text of those sections of Title 35, U.S. Code not included in this action can be found in a prior Office action.
Claim(s) 1, 7-9, 15, and 16 is/are rejected under 35 U.S.C. 103 as being unpatentable over Koneru et al. (US Patent 12,130,848) in view of Choi et al. (US PAP 2025/0390318).
Consider claim 1, Koneru teaches a computer-implemented method (abstract) comprising:
receiving, at a monitoring agent layer, a user prompt from a client device (col 16 line 1-15) associated with a user via a channel from plurality of connected channels (col 4 line 30-40 channels used to receive user prompts and send responses), the monitoring agent layer monitoring prompt characteristics associated with the user prompt prior to providing the user prompt to an orchestrator agent layer (col 15 line 55- col 20, associated data record, see col 8 lines 49-65, data records include things like transcript of previous interactions);
processing, by the orchestrator agent layer, the user prompt to generate one or more task prompts from the user prompt (col 16 line 10-26, sending prompt to orchestrator to determine execution instruction);
providing the one or more task prompts to a pre-trained large language model to generate one or more task responses based on the one or more task prompts, wherein the pre-trained large language model comprises one or more fine-tuned layers (col 16 lines 27-45, providing execution instruction to NLU LLM. Col 7 lines 54-65, NLU LLM is an application specific LLM fine-tuned on task.);
orchestrating, based on the orchestrator agent layer receiving the one or more task responses, one or more task items by providing a first task item instruction to an internal platform agent or a second task item instruction to a third-party platform agent (col 16 line 40-65, orchestrator LLM is provided response from NLU LLM, and determines that API for external resource should be queried, which is then queried); and
generating a user response to the user prompt according to a task item status received from the internal platform agent or the third-party platform agent (col 17 line 7-col 18 line 6, generating a response based on API return.).
Koneru does not specifically teach receiving a confirmation of the one or more task responses based on providing the one or more task responses for display on a graphical user interface of the client device;
orchestrating… based on receiving the confirmation.
In the same field of orchestrating user request, Choi teaches receiving a confirmation of the one or more task responses based on providing the one or more task responses for display on a graphical user interface of the client device (0142-47, prompting user via display for confirmation and receiving confirmation);
orchestrating… based on receiving the confirmation (0142-47, once user confirmation received, API request may be sent.).
It would have been obvious to one of ordinary skill in the art at the time of effective filing to seek user confirmation as taught by Choi in the system of Koneru in order to prevent erroneous commands from being executed.
Consider claim 7, Koneru teaches The computer-implemented method of claim 1, further comprising:
providing the one or more task prompts to one or more adapters (col 6 lines 20-26, conversation management framework receives response from orchestrator LLM);
modifying, by the one or more adapters, the one or more task prompts from a first format to a second format (col 6 lines 20-26, conversation management framework converts response to query for NLU LLM); and
inputting the second format of the one or more task prompts to the pre-trained large language model (col 16 lines 27-30, providing prompt to NLU LLM).
Consider claim 8, Koneru teaches The computer-implemented method of claim 1, further comprising:
providing the one or more task responses to one or more adapters (col 16 lines 38-41, conversation management framework receives output from NLU LLM );
modifying, by the one or more adapters, the one or more task responses from a first format to a second format (Col 16 lines 41-50, conversation management framework converts received to another prompt for the orchestrator LLM); and
providing the second format of the one or more task responses to the orchestrator agent layer (Col 16 lines 41-50, conversation management framework sends prompt to orchestrator LLM).
Consider claim 9, Koneru teaches A system (abstract) comprising:
at least one processor (col 6 lines 37-47, processors); and
at least one non-transitory computer-readable storage medium storing instructions that, when executed by the at least one processor (col 6 lines 48-67, memory, RAM/ROM etc), cause the system to:
receive, at a monitoring agent layer, a user prompt from a client device (col 16 line 1-15) associated with a user via a channel from plurality of connected channels (col 4 line 30-40 channels used to receive user prompts and send responses), the monitoring agent layer monitoring prompt characteristics associated with the user prompt prior to providing the user prompt to an orchestrator agent layer (col 15 line 55- col 20, associated data record, see col 8 lines 49-65, data records include things like transcript of previous interactions);
process, by the orchestrator agent layer, the user prompt to generate one or more task prompts from the user prompt (col 16 line 10-26, sending prompt to orchestrator to determine execution instruction);
provide the one or more task prompts to a pre-trained large language model to generate one or more task responses based on the one or more task prompts, wherein the pre-trained large language model comprises one or more fine-tuned layers (col 16 lines 27-45, providing execution instruction to NLU LLM. Col 7 lines 54-65, NLU LLM is an application specific LLM fine-tuned on task.);
orchestrate, based on the orchestrator agent layer receiving the one or more task responses, one or more task items by providing a first task item instruction to an internal platform agent or a second task item instruction to a third-party platform agent (col 16 line 40-65, orchestrator LLM is provided response from NLU LLM, and determines that API for external resource should be queried, which is then queried); and
generate a user response to the user prompt according to a task item status received from the internal platform agent or the third-party platform agent (col 17 line 7-col 18 line 6, generating a response based on API return.).
Koneru does not specifically teach receiving a confirmation of the one or more task responses based on providing the one or more task responses for display on a graphical user interface of the client device;
orchestrating… based on receiving the confirmation.
In the same field of orchestrating user request, Choi teaches receiving a confirmation of the one or more task responses based on providing the one or more task responses for display on a graphical user interface of the client device (0142-47, prompting user via display for confirmation and receiving confirmation);
orchestrating… based on receiving the confirmation (0142-47, once user confirmation received, API request may be sent.).
It would have been obvious to one of ordinary skill in the art at the time of effective filing to seek user confirmation as taught by Choi in the system of Koneru in order to prevent erroneous commands from being executed.
Claim 15 contains similar limitations as claim 7 and is therefore rejected for the same reasons.
Consider claim 16, Koneru teaches a non-transitory computer-readable medium storing instructions thereon that (col 6 lines 48-67, memory, RAM/ROM etc), when executed by at least one processor, cause a computing device to:
receive, at a monitoring agent layer, a user prompt from a client device (col 16 line 1-15) associated with a user via a channel from plurality of connected channels (col 4 line 30-40 channels used to receive user prompts and send responses), the monitoring agent layer monitoring prompt characteristics associated with the user prompt prior to providing the user prompt to an orchestrator agent layer (col 15 line 55- col 20, associated data record, see col 8 lines 49-65, data records include things like transcript of previous interactions);
process, by the orchestrator agent layer, the user prompt to generate one or more task prompts from the user prompt (col 16 line 10-26, sending prompt to orchestrator to determine execution instruction);
provide the one or more task prompts to a pre-trained large language model to generate one or more task responses based on the one or more task prompts, wherein the pre-trained large language model comprises one or more fine-tuned layers (col 16 lines 27-45, providing execution instruction to NLU LLM. Col 7 lines 54-65, NLU LLM is an application specific LLM fine-tuned on task.);
orchestrate, based on the orchestrator agent layer receiving the one or more task responses, one or more task items by providing a first task item instruction to an internal platform agent or a second task item instruction to a third-party platform agent (col 16 line 40-65, orchestrator LLM is provided response from NLU LLM, and determines that API for external resource should be queried, which is then queried); and
generate a user response to the user prompt according to a task item status received from the internal platform agent or the third-party platform agent (col 17 line 7-col 18 line 6, generating a response based on API return.).
Koneru does not specifically teach receiving a confirmation of the one or more task responses based on providing the one or more task responses for display on a graphical user interface of the client device;
orchestrating… based on receiving the confirmation.
In the same field of orchestrating user request, Choi teaches receiving a confirmation of the one or more task responses based on providing the one or more task responses for display on a graphical user interface of the client device (0142-47, prompting user via display for confirmation and receiving confirmation);
orchestrating… based on receiving the confirmation (0142-47, once user confirmation received, API request may be sent.).
It would have been obvious to one of ordinary skill in the art at the time of effective filing to seek user confirmation as taught by Choi in the system of Koneru in order to prevent erroneous commands from being executed.
Claim(s) 2, 3, 10, 11, 17, and 18 is/are rejected under 35 U.S.C. 103 as being unpatentable over Koneru and Choi as applied to claims 1, 9, and 16, and further in view of Hayes (US Patent 12,417,359 ).
Consider claim 2, Koneru and Choi teach the computer-implemented method of claim 1, but do not specifically teach transforming, by the monitoring agent layer, the user prompt based on one or more security protocols.
In the same field of LLM prompting, Hayes teaches transforming, by the monitoring agent layer, the user prompt based on one or more security protocols (col 8 line 53- col 9 line 16, user prompted may be filtered to removed harmful content, etc).
It would have been obvious to one of ordinary skill in the art at the time of effective filing to filter user prompts as taught by Hayes in in the system of Koneru and Choi in order to maintain the security features of the LLM (Hayes Background).
Consider claim 3, Koneru and Choi teach the computer-implemented method of claim 1, but do not specifically teach filtering the user response according to one or more security protocols prior to providing the user response to the client device associated with the user.
In the same field of LLM prompting, Hayes teaches filtering the user response according to one or more security protocols prior to providing the user response to the client device associated with the user (col 9 line 29-51, output from LLM may be filtered for harmful content).
It would have been obvious to one of ordinary skill in the art at the time of effective filing to filter outputs as taught by Hayes in in the system of Koneru and Choi in order to maintain the security features of the LLM (Hayes Background).
Claim 10 contains similar limitations as claim 2 and is therefore rejected for the same reasons.
Claim 11 contains similar limitations as claim 3 and is therefore rejected for the same reasons.
Claim 17 contains similar limitations as claim 2 and is therefore rejected for the same reasons.
Claim 18 contains similar limitations as claim 3 and is therefore rejected for the same reasons.
Claim(s) 4, 5, 12, 13, and 19 is/are rejected under 35 U.S.C. 103 as being unpatentable over Koneru and Choi as applied to claims 1, 9, and 16, and further in view of Bao et al. (US PAP 2025/0259019).
Consider claim 4, Koneru and Choi teach the computer-implemented method of claim 1, but do not specifically teach generating the one or more fine-tuned layers of the pre-trained large language model, wherein the one or more fine-tuned layers include at least one of: a demographics layer, an industry layer, or a domain layer.
In the same field of fine-tuning LLMs, Bao teaches generating the one or more fine-tuned layers of the pre-trained large language model, wherein the one or more fine-tuned layers include at least one of: a demographics layer, an industry layer, or a domain layer (0017-18 fine-tuning LLMs with domain knowledge, figure 3 adapter layers).
It would have been obvious to one of ordinary skill in the art at the time of effective filing to use domain knowledge to fine-tune the LLM as taught by Bao in the system of Koneru and Choi in order to enhance the performance of the LLM for specific domain tasks.
Consider claim 5, Bao teaches The computer-implemented method of claim 4, further comprising generating the one or more fine-tuned layers according to a knowledge graph (0017-18 fine-tuning LLMs with domain knowledge, figure 3 adapter layers).
Claim 12 contains similar limitations as claim 4 and is therefore rejected for the same reasons.
Claim 13 contains similar limitations as claim 5 and is therefore rejected for the same reasons.
Claim 19 contains similar limitations as claim 4 and is therefore rejected for the same reasons.
Claim(s) 6, 14, and 20 is/are rejected under 35 U.S.C. 103 as being unpatentable over Koneru and Choi as applied to claims 1, 9, and 16, and further in view of Buchanan et al. (US PAP 2025/0298988).
Consider claim 6, Koneru and Choi teach the computer-implemented method of claim 1, but do not specifically teach:
determining, by the monitoring agent layer, an escalation event associated with a user based on monitoring the prompt characteristics associated with the user prompt; and
performing a de-escalating action according to the escalation event.
In the same field of LLM chat engines, Buchanan teaches determining, by the monitoring agent layer, an escalation event associated with a user based on monitoring the prompt characteristics associated with the user prompt (0070 recognizing aggressive language etc); and
performing a de-escalating action according to the escalation event (0070 performing de-escalation).
It would have been obvious to one of ordinary skill in the art at the time of effective filing to include de-escalation as taught by Buchanan in the system of Koneru and Choi in order to enhance the user experience in the conversational system.
Claim 14 contains similar limitations as claim 6 and is therefore rejected for the same reasons.
Claim 20 contains similar limitations as claim 6 and is therefore rejected for the same reasons.
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to DOUGLAS C GODBOLD whose telephone number is (571)270-1451. The examiner can normally be reached 6:30am-5pm Monday-Thursday.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Andrew Flanders can be reached at (571)272-7516. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
DOUGLAS GODBOLD
Examiner
Art Unit 2655
/DOUGLAS GODBOLD/Primary Examiner, Art Unit 2655
Prosecution Insights